It seems to me that this is just recreating the incentives that exist now for pundits--if you say what everyone else is saying, you're a sensible fellow who gets to keep being heard from; if you disagree with everyone else, you're some kinda nutcase who shouldn't be heard from anymore.
>I think the advantage of iterating it this way is that you can amplify small changes. Suppose that right now, the market thinks there’s a 50% chance that America will beat China in 2100. But I am a great economic analyst who knows things the market doesn’t, which allow me to determine that the real chance is 52%. Leaving money in a market for five years to make 4% return doesn’t sound great. So we could do something like make people bet on whether the 2025 market would be <40%, 40 - 45%, 45-50%, 50-55%, 55-60%, or >60%. This way if you’re even directionally right you can double your money in five years. I bet there are more clever mathematical ways to do this which would give you finer-grained resolution.
Essentially these are stock options with fairly high premiums dated anywhere from one to three years. These exactly solve your issue of amplifying small changes, as the price of an existing options contract automatically reflects changes in market consensus.
As an example, a Ford 01/24 20C will cost you $4.8 https://www.nasdaq.com/market-activity/stocks/f/option-chain in premium, while a 01/23 20C runs you $3.65. If you think Ford will go up more than decay by, say, 06/22 (fairly low over these time periods) you can buy one of these options and, if Ford goes up, sell for more than you bought it for whenever you want to, regardless of the actual expiry date.
What Scott is really talking about is best represented by an option fence https://www.investopedia.com/terms/f/fence-options.asp. Its and option strategy such that it pays out if an underlying security is within and range but pays nothing outside it.
With a liquid enough market in options with a wide range of strikes you can essentially define the market's view of the probability that an asset being in any range at a given point in time.
If you make the fence size small enough you approximate an Arrow-Debreu security which is a contract that only pays out in one state of the world.
There's the possibility for a party-destroying primary in 2024 if Biden walks away. Harris is "in line" and is a treasured symbol to the activist identitarian wing that has taken over so much of the party; she's also one of the least popular politicians in the country. If she's nominated she'll likely be crushed by Ron Desantis or whoever, and it would essentially be a do-over of 2016 - nominating a deeply unpopular candidate who is constantly defended along identitarian lines. But if the opposition coalesces around Buttigieg then you'll get the Bernie Bro vs Hillary Stan war again, only much harsher. Not great!
"An election or two" was your timeframe. And while it's hardly a settled issue within Marxism, Marxists have participated in bourgeois horserace politics for a long time.
There are certainly scenarios in which the Democratic Party in its current form ceases to exist. #1 is nuclear war, since political affiliation of the dead would massively slew Democratic and also they're typically in favour of cutting military spending which would discredit them. #2 is that the Democrats do something treasonous on the level of the Confederacy and get consigned to the political black hole in the aftermath (I know they weren't after the literal Confederacy, but AIUI it wasn't too far from happening).
I'm having a hard time seeing a line from the 2024 Democratic primary to #1, though. #2 seems slightly more plausible, but still far from clear.
Communists have run for office across Europe, with much success in the past. Particularly in France and Germany. The Die Linke party in Germany is a successor to the East German communists, and the Russian communists are the second party in Russia.
Marxists who don’t want to engage in bourgeois politics tend to be tiny sectarians, or entirely online.
Don’t they? I know Marxists in France are anti-Identitarian. I was responding to your jive about “bourgeois horse race politics” (elections)-communists do engage in that. And in the US too, with little success.
US online Marxists seem to lack any organisational power whatsoever.
> a treasured symbol to the activist identitarian wing
Is she? Basically all of the leftists I know, both online and in person, hate her because of her prosecutorial background and the fact that she’s fairly centrist. I’d predict that they would coalesce around Elizabeth Warren more than around her, though of course she’s in a great position to start with because she’s the VP.
Yeah in the primary she topped out at what, 5%? It doesn't seem like there's much of a constituency that's particularly attached to her. Being VP will help but I'm not sure it's enough to lead to the situation Freddie is describing.
I don't think many of the people who supported Biden in the last primary were particularly "attached" to him either. But he had plenty of establishment backing, which made him the default option. It feels like Harris is going to receive that establishment backing in 2024, and many of those same types of people will support her because of that, if not explicitly.
All the leftists I know, both online and in person, also hate her for her prosecutorial background and centrist economic tendencies, but I don't presume to think that the leftists I know are representative of democratic voters. The leftists I know mainly supported Bernie, with some going for Warren, or possibly Yang. Clearly they weren't representative last time. As of right now, if Harris wants the nomination, it seems like it's hers, unfortunately. Unless the field is small and progressives can unite behind another non-establishment figure.
My point wasn't that there's no way she could win, it's that if she runs and loses in a close race I wouldn't expect her supporters to be sore losers in the way Freddie was implying. That kind of support seems more likely to coalesce around whoever her opposition ends up being, although I have trouble seeing it for Buttigieg either- the "activist identitarian wing" didn't seem to like him much either. Ocasio-Cortez seems most plausible but who knows if she even runs.
Gotcha. I think people are using the terms "activist identitarian wing" and "leftist" differently here. I would not describe Bernie as an identitarian. To me, AOC fits the bill, but she's also got a good amount of support from the "economically progressive non-identitarian wing" - closer to what I consider "leftists". It seems to me like conservatives tend to conflate these two camps. Harris is obviously an activist identitarian, but I also think that what I consider to be the activist identitarian wing views Buttigieg favorably. As such, I think a head-to-head between Harris and Buttigieg would be very different from the Bernie vs Hillary war. Not much ideological difference there.
Hard to predict who will represent the "leftist" camp in the 2024 primary. AOC is the main person who comes to mind, but I'm pretty sure she'd get destroyed in a general election. Pramila Jayapal and Sherrod Brown might be better options.
I need to start checking Metaculus more. I was hoping to find something about Rittenhouse on Polymarket, but it didn't come up. Watching the trial has been rough. I'm not looking forward to the news following the results of the trial.
The prosecution had their work cut out for them given that pretty much the entire event was captured on video and clearly supported the self-defence argument.
The prosecutor was reduced to claiming that the defendant's desire to protect himself was unreasonable which strikes me as a tough sell.
There were other things they could have done; more charges than murder or negligent homicide.
I'm not a lawyer and I've only done enough research to not accidentally commit a crime with my guns, but I wouldn't have done what he did.
He's a minor, he crossed state lines with a gun, the gun was being carried illegally (Fuck off with that hunting bullshit, the judge himself was fine with that charge until the cameras were rolling), Brandishing, it's every crime you can commit by having a gun except for stealing it in the first place or concealing it.
The thing I have trouble with is attempting to disprove the self defense claim, because I don't actually know anything about what qualifies you. My instinct (and the lectures I got about it) tell me this shouldn't be a covered offence.
If I drove across state lines to the Texas Republican Convention and stood their strapped and shittalking until someone did something and then shot them, that is NOT self defense.
He carried the gun legally, that's clear. Crossing state lines means nothing, it was a 15 minute drive anyway. Brandishing maybe- not sure what you mean by that. Just carrying an assault weapon around strapped to your body is not brandishing, in many states that is perfectly legal and people do it. If he was pointing his gun at people that may be an issue and the prosecution tried to make that argument.
There is no evidence of shit-talking that I'm aware of, but I think that could be relevant depending on what exactly you said. Of course the guy he shot was screaming that he was going to kill Rittenhouse and eat his heart and stuff like that...
That's exactly the opposite of clear, which is the point of the judge's actual ruling. Even as a professional with decades of experience in criminal law and many hours studying that particular statute, he professed uncertainty as to what was or was not legal.
*Because* it is unclear, Rittenhouse cannot be prosecuted for it. That's pretty solid in Anglo-American criminal law. But "you can't be prosecuted for X" and "X is legal" are not the same thing.
Essentially I think that what's happening is that for the group of randos in the first experiment, "predict reality", and "predict what superforecasters will predict" don't lead to different predictions. So """"reciprocal scoring"""" performs well compared to scoring against reality for the group of randos. But note that """"reciprocal scoring"""" in the first experiment is not actually reciprocal; the superforecasters get scored against reality, not against the randos.
The lesson I'd want to teach here is "it is impossible to completely predict something that can read your prediction before it does the predicted thing".
(This is part of probably the most important computer-science proof ever - the proof that the halting problem is unsolvable. The proof amounts to: suppose there's an algorithm X that for any input algorithm Y returns, in finite time, "yes" if Y halts in finite time and "no" if it runs forever. Now I make an algorithm Y that goes "feed Y's own source code into X, and do the opposite of what it predicts". By assumption X finishes in finite time for any Y, so Y will always be able to finish in finite time if X predicts it runs forever, and certainly it can spin the wheels forever if X predicts it finishes in finite time. But this is a counterexample to the definition of X, so X doesn't exist.)
This is a good description of my objections to Newcomb's problem. The existence of the perfect predictor and its prediction power is supposedly a given, but I'm not convinced it's even mathematically coherent, so it may be intractable to actually reason mathematically about.
A version of Newcomb's paradox in which the boxes were transparent would definitely have this problem, but the normal version doesn't because you don't have access to the predictor's prediction before you act. You could, in theory, physically enforce this; you put the predictor on Mars and run it ~simultaneous with the actual choice (so that light-lag forces the events to be independent) and, instead of having physical objects in the boxes, just agree to the payoff matrix of Newcomb's paradox based on the prediction and on your action.
Yes, I had exactly this debate on reddit very recently, which I summed up as:
I think this gets to the ultimate point though, which is that Newcomb's paradox is not a single problem but a family of problems whose answers are very sensitive to the specific constraints. If you specify a perfect predictor, then the question is inconsistent. If you reduce the predictor's accuracy to make it consistent, you now open possibilities for two-box solutions. You simply cannot close those loop holes via contrivances on the inputs, because all that does is entail complementary contrivances in the solutions.
So like I said initially, on average you're better off picking one box, but there simply must exist circumstances in which two boxes is the correct answer too. That is, such solutions simply must exist by logical necessity.
> If you reduce the predictor's accuracy to make it consistent, you now open possibilities for two-box solutions.
How so? The payoff matrix doesn't meaningfully change with arbitrarily high accuracy even if it's short of perfection, so the two-boxer is still expected to walk away with significantly less money.
"Arbitrarily high accuracy" is just as impossible as perfection.
Regardless, what you're looking at here is akin to zero-determinant strategies for the Prisoner's Dillemma. If you have any knowledge of the predictor's history and/or its internal workings, you have some non-zero chance of winning the two box solution.
> "Arbitrarily high accuracy" is just as impossible as perfection.
The original formulation gives boxes with the values of $1,000,000 and $1,000. Exactly what do you think is the maximum accuracy prediction possible, and how much does it change the math?
There *are* strategies where the player introduces randomness into their own decision, and this does indeed lower the predictor's accuracy correspondingly. Do any of these improve the player's expected winnings?
>If you have any knowledge of the predictor's history and/or its internal workings, you have some non-zero chance of winning the two box solution.
"Winning" is not defined as being theoretically able to choose to two-box and walk away with $1,001,000. If that non-zero chance is sufficiently close to zero, two-boxing is still a bad idea just as surely as betting on 00 at a roulette table.
Newcomb's paradox has no issue with the halting problem so long as you can't see Omega's prediction. Its prediction doesn't impact your decision at all, so it can be correct. If you can see what prediction Omega makes, than Omega is solving the halting problem unless you happen to make the same decision regardless. Maybe Omega makes predictions for what everyone passing by would do in each scenario, and only opens the door for those who's actions wouldn't be inconsistent?
> Newcomb's paradox has no issue with the halting problem so long as you can't see Omega's prediction
This isn't correct. The Halting problem isn't undecidable just because the program can change its behaviour based on the prediction, that's just the simplest degenerate case. Goedel proved that there are in fact an infinite number of undecidable propositions. Undecidability is fundamental and inescapable.
Since a human can simulate a Turing machine, then Omega must have its own set of undecidable propositions for which predictions simply must be considerably less than certain.
"Since a human can simulate a Turing machine" [citation needed]
But it's easy to get around that. Just add a provision that says you need to choose within 10 seconds, or else you automatically 1 box it. Though realistically, unless you never choose a box due to an unterminated loop or something, the halting problem shouldn't enter into it anyway.
And the distinction between "good forecaster" vs "academic" vs "pundit" seems important too - so-called "experts" are quite often political animals rather than truth-seekers.
Wait, I was just wondering if this considering this problem was how Robin Hanson came to think of futarchy. Doesn't the commitment to picking the choice predicted to be better turn it into causal decision theory?
I don't think so. In the example, policymakers are trading off the benefits of a mask mandate against other values. The benefits are larger when the pandemic is very bad, so even if you commit to picking the choice predicted to be better you are still more likely to implement mask mandates if the pandemic is bad. So the same problem occurs.
I'm talking about a changed legal environment where what we now call the authorities have no choice of policy. Their job is to implement the decision which the prediction market made. And this is common knowledge of the participants in the market. https://en.wikipedia.org/wiki/Futarchy
Yeah that's what I'm talking about too, that's what I meant when I said "if you commit to picking the choice predicted to be better". Maybe you thought I was talking about a scenario where politicians say that they are going to follow the prediction markets, but still have the ability to do something different?
The opposite. I model futarchy as: the market estimates (Y1 deaths | do(mask mandate)) and (Y2 deaths | do(no mask mandate)); then the state mandates masks iff expectation(Y1) < expectation(Y2). In this model the decision node (mask mandate? or not?) is screened off from any nodes about the *reasons* for the decision. The betters will have a variety of models of the situation but they'd all share this property (assuming they trust the institutions to work as claimed).
Actually, the state will mandate masks if the difference in deaths is more than the costs of the mask mandate policy in terms of other variables that the state has pre-defined. If the pandemic is bad then the deaths saved by the mandate are more likely to overcome this threshold, even if the mandate saves lives regardless of how bad the pandemic is. So the state is more likely to implement a mask mandate if the pandemic is bad.
The networks basically have three classifications (in a 2-person race): Called for A, Called for B, Too Close to Call. They don't want to deal with things like "favored", and given what we saw in the 2020 presidential election, I'm inclined to agree with them. The NBC and ABC decision desks won't call a race unless they're at least 99.5 confident in the winner. PredictIt and Polymarket weren't at that point yet, although prediction markets can be a little wonky around the extremes.
Yeah Scott seems to be missing the fact that in this context "call" is a technical term meaning announcing that a candidate has won, not just saying "this candidate seems to be winning." 98% just *is* actually too close to call by this standard.
But you're not giving people log(101) bits of information. Information is not conveyed magically from the mind of the reporter to the mind of the reader.
Treat the news as a noisy channel, with reduced information capacity. Headlines magnify that problem since informational (non-clickbait) headlines try to convey information without the benefit of the full article.
I mean reporting a percentage 0-100%, like the betting markets do, is lg(101) bits of information. News reports compress that range down to just three options: Yes, no, and "too close to call". I'd be happy if the news just quoted the prediction markets.
In particular, extrapolate it to a 50-state presidential election. If you call each state when it's 98% sure, your chance of getting them all right is .98^50 == .36, so 64% chance at least one is wrong. Let's be more realistic: plenty of states never get as low as 98%. Keep it to the 10 or so "battleground states": now 98% right gets you "only" a 18% chance of getting one wrong. Hope that one wasn't Florida 2000!
In addition, if you read election day articles, or watch election day TV news coverage, the analysts often say things like "Candidate A is favored", or "Candidate B has an increasingly narrow path to victory. These aren't exactly probabilities but tell the same story as them.
This is the correct answer. See also the Fox (and eventually AP) call of Arizona in 2020, which was predictably correct at the time it was made but nowhere near the 1/200 threshold it should have been and therefore a bad call.
Scary detail in that "Expert forecasts vs. reality" tweet that's easy to miss (and complicates the story of systematic mis-calibration a bit): up through 2008 or so the life expectancy projections consistently erred on the low side. Since then they've consistently erred on the high side.
Related: When one sees these instances of experts repeatedly predicting a trend change and repeatedly being wrong, one is inclined to lambast them saying "They should have seen all these previous predictions of a similar trend change, and how they had always been wrong, so maybe they should have learned the lesson and they should have just predicted that the trend would continue." But in what time frame should they look at the trend? After 2008, should they predict that the long-term trend (one of relatively fast increasing life expectancy) will continue, or the short-term trend (of much slower increase) will continue? Or maybe they should predict that the short-term will continue *on the short term* (but neither prediction is obviously right on the long term)?
In fairness to traditional media, part of the issue is that their standards of proof are super high. They make hundreds of predictions per election, and they get one wrong maybe once every decade or so. They probably do know that the chance of a particular candidate winning is 98% (or whatever), but by their standard that just isn’t good enough, because that would lead to many more missed calls.
That's the idealistic take. The cynical take is that they just want you to keep watching, and as soon as they call it one way or the other, half the audience is going to change channels.
Honestly it’s more realistic to say that they just are really conservative on calling it. If they get even one major call wrong they’ll get lambasted for weeks, and no one is materially better off for knowing the result a couple hours sooner.
I have seen the third graphic of XiXiDu from people who claimed that the graphic is nonsense and that XiXiDu got it completely wrong. The many horizontal lines are not predictions. Rather they are summed up plans of governments of various countries on how much solar capacity the governments plan to fund in the next years.
I didn't verify the claim, so if someone can say something about it (or about the other three graphs), that would be nice!
What does the prediction markets say about Trump running for - and winning - the Republican nomination for US president in 2024?
I ask, since back in 2020 I betted a bottle of good italian wine plus a dinner at a nice italian restaurant that he will run & get the nomination in 2024. (Hedging, really)
More generally, I am worried that if none of the sane US political people pick up obviously sensible policy proposals, like immigration control of irregulars & compensating those who lose out of economic globalisation, then a window opens for the insane to take power. Like in Russia 1917 and Germany 1934. Not as dramatic as there, but still.
May sanity prevail among the not-insane in the US. It is important also for Europeans, and everyone else for that matter.
PredictIt shows a 40% probability that Trump will win the Republican nomination. But this far out, I wouldn't give it a lot of credence. Primary elections are hard to forecast because the candidates are so similar ideologically.
My own subjective estimates, for what they are worth, have been 50/50 that he will run, 70/30 that he will win the Republican nomination, and 40/60 that he will win the election. Which gives a probability of .5x.7=.35 that he will win the nomination (enough to bet a bottle of good wine), but only .5x.7x.4=.14 that he will become the next US president. (For Trump himself it would boil down to .7x.4= .28 though.)
All subjective estimates are basically talking-through-one's-hat (as the English say), but on the other hand you often need distance to see things clearly. I never understood my own country better than after having lived a year in the US (the fog lifted after approx. six months), and maybe it works the other way as well.
Seriously, all the havering from the Democratic side about Trump makes me think they are building him up as a credible candidate in 2024.
While it's to their short-term advantage to keep scaremongering about the Trumpian menace and likely coups, in the long-term they are contradicting themselves about him being powerless and ineffectual and reduced to angry outbursts on social media from his exile in Mar-a-Lago when they and their supporters write lurid pieces about plotting coups:
We are in the middle of the defining-the-situation political game (which is at least 80 percent of what media-led political discussions are about, everywhere), and it is always dangerous to run two conflicting narratives/definitions of the situation at the same time (Trump is cunning and clever/Trump is a nutcase).
I have a hard time making up my mind myself. Sometimes I believe he is both, which is the most dangerous kind of politician. I actually hope he is ”only” cunning and clever, since I would much prefer a clever half-crook to occupy the White House (like Nixon), than somebody who - insanely - really believe that the election was stolen from him (plus the old Obama birth certificate nonsense).
With the benefit of hindsight, claiming “the election was stolen from me” is rather clever, if he plans to run again in 2024. The US dislikes a loser. Running like “ last time’s loser tries a second time” is unlikely to cut it. Running like “last time’s winner whose victory was stolen from him tries a second time” is a better narrative, from Trump’s point of view. Not much better perhaps, since he must convince the median voter (who hopefully prefer a sane ruler) that he is not insane enough to actually believe so himself, and only pretends to believe this, because he is very clever and knows it positions him better for a re-run…the message is rather convoluted.
However, if the Democrats mess things up sufficiently, with lack of immigration control, no protection of the US working class against competition from abroad, the likely very costly “green shift”, and fumbled foreign policy, enough voters may hold their nose and take a chance with him again in 2024.
A better strategy for the Democrats would be to instead "steal" many of Trump's policies in the years ahead (packaged a bit differently, of course). If they fail to do this, I’d adjust my prior of him winning in 2024 to 50/50 (if he runs)😊.
But isn't Trump actually great for them, compared with somebody who is both opposed to them and might have the ability to do something about it? Trump had the majority in both Houses for two years and the only thing he managed to do was slashing some taxes, the apocalyptic-level menace that he is.
If Biden does no better on COVID (and despite a vaccine it looks pretty similar to me - the restrictions are still in place; more people have died of COVID with Biden in office than Trump), and there are other things that bother people (inflation, Afghanistan, immigration, whatever) that is still an issue approaching 2024, then I think a lot of people will remember the Trump years differently than they would have said at the time. Biden ran as a centrist uniter, but hasn't really governed as such. Trump lost in large part because of COVID, otherwise the very close election could have easily, I would say would have, gone Trump's way.
Trump's years, outside of the Russia allegations and the tweets/self-aggrandizing lies, were actually pretty calm and prosperous. With the Russia allegations looking worse for the media sharing them, Trump's personality and weird lies are the last major issue. People tend to forget over time, or remember more fondly, concerns of the past. With your point that Trump didn't change a whole lot, it may sound to many like a fine alternative in practice, even if Trump himself looks like/is a mess.
Trump is less likely to win than a moderate Republican, which means he's a better opponent to have in that sense. But the downsides of him winning are larger. In particular if you are concerned about the erosion of democracy, a small chance of never being able to win an election again is worse than a higher chance of losing one now
Sure, if those are the two options. What the establishment is actually afraid of, it seems to me, is that a much more competent Trump might appear, a populist authoritarian who has a realistic chance to bring about that erosion. If there's an actual demand for this sort of personality on the right, a prominent buffoon already in place seems to be a lesser risk than a vacancy which might get filled by an unpredictable dark horse.
To be honest, I don't see how reciprocal forecasting is supposed to work in game-theoretic sense. Normal prediction market has nice and obvious Nash equilibrium: just predict the probability as best you can. Not so in reciprocal forecasting. From what I can tell, a strategy closest to optimum is to shout at loud as you can "Let's all vote for X" and then vote for X. If it is even possible to do this without communicating with others by forecasting the most trivial outcome, like answer 50% to any question.
Iterative prediction is far more plausible. You don't even need to do the iteration if you ensure two conditions: 1. The stakes are freely tradable at any time. 2. While the prediction stays the same, the value of the stake is growing at least as fast as some default assets like S&P 500. If the second condition is not fulfilled, then unless the market is extremely wrong, long-term it is more profitable to invest in other asserts. This can be achieved by, you guessed it, having the market operator invest the stakes in that very investment product. This could even happen automatically by making the stakes in some cryptocurrency like ETH, if you believe that ETH will continue growing at least as fast as S&P 500.
2) is possible to achieve by availability of leverage. For example, for any sufficiently liquid base contract, one can set up a futures market with risk based margining. Then each long or short participant can invest most of their cash in any asset mix they like at each point in time. This may be easier to implement than to mitigate credit risk on a market investing market operator or agreeing on a single default asset.
Biden himself has said some things about being a one term president. Mixed with his age and his low approval ratings, it's not surprising that the markets are looking for alternatives.
It's a sad state of affairs that the candidates available are as poor as they are. I don't think any of the listed candidates can beat Trump, let alone a more solid Republican that doesn't make half the country reflexively convulse at hearing their name. Those that excite the base scare the middle, and vice versa. The fact that Hillary Clinton appears in the group is certainly not a great sign for Democrats. It's early yet, though, and maybe some new candidates or a different consensus candidate can still emerge. Or again, maybe Biden turns things around and doesn't get any health issues and decides to run again. Obviously I'm not expecting that to all happen. Even if he does turn around his approval ratings, that might be a great time to retire with a good track record and leave a legacy, rather than chancing that he'll make it to 85 in one of the most grueling jobs in the world.
Biden would likely lose right now to most Republican candidates, though he has a few years to potentially turn around his unpopularity. Harris is riding her own unpopularity as well as Biden's. Interestingly enough is the recent soft takedown articles from left-leaning mainstream news sources. They don't seem to like her either, and may be paving the way to have another candidate in 2024. Buttigieg is not popular with the farther left, and most of the others are multi-time losers who are even older than they were the last time they ran. AOC is too inexperienced to run a nationwide campaign, and too limited in her popularity to unite the Democrats. She's a darling of the far left and young people, but unless she comes sharply to the center, she's going to scare away a lot of the middle and independents.
I read that list as Democrats being in real trouble in 2024. Trump could be divisive enough to give some of them a chance, maybe, but a better candidate seems like a pretty obvious win for Republicans at the moment.
I think a lot of that list represents unknown unknowns - a similar list in 2013 for the 2016 election wouldn’t have listed Donald Trump at all!
People think that Biden won’t run and Kamala won’t win, so they want to sell Biden and Kamala and without any other reasonable options yet they just list famous Democrats. I don’t think there’s a chance that AOC will make a serious bid, Bernie won’t run because of his age, and Clinton definitely won’t run either.
Buttigieg and Warren are both realistic options (and correspondingly get some predictions on the list) but it’s so far away from the election that it’s still pretty meaningless.
Buttigieg maybe, in time - he's currently Secretary for Transport so he is getting that experience of being in government under his belt, but he's still young and inexperienced (for a politician) and he's not wired into a political machine like Obama and Chicago.
I think Warren has shot her bolt - she had a chance last time, but since Biden was always going to be the pick that went nowhere, and she kind of has "a great future behind her" - she was touted as going to do great things, but she's been around now for a while and that kind of candidacy never really did happen. I think, to be frank, that by 2024 she will be too stale.
I have no idea who a potential Democratic candidate that is not Harris or Biden might be, though.
"a similar list in 2013 for the 2016 election wouldn’t have listed Donald Trump at all!"
Yes, the most amusing thing that came out of the 2016 result were the victory videos showing all the pundits confidently predicting "Clinton will win this state, she will take that state, the polls say such-and-such" and on the night, those states went for Trump.
I think everybody enjoys seeing pundits ending up with egg on their faces. The sniffy certitude on display that of course Hillary had it in the bag makes it all the sweeter.
As well as everyone from Lin-Manuel Miranda up to Obama saying "Never gonna happen" and then it did.
I definitely agree that it's too early to rule out other, new candidates. The problem is Biden sucking up the official oxygen by being a first term president. To actively put themselves out there when he hasn't said whether he plans to run again risks going against the party. If he does decide to run, and presuming he hasn't fixed his approval ratings sufficiently, then potential candidates are going to have to really think about whether to run anyway, instead of sitting things out. Tough choice all around, unless he backs off and makes way.
If it does turn out to be Biden versus Trump in 2024, I have no idea what will happen.
You would expect that selecting Kamala Harris as Vice President was giving her the chance to build her candidacy for 2024, but there are some odd little stories here and there which whisper at the rift within the lute and dissatisfaction (allegedly) with her performance to date:
It's funny you say Hillary Clinton being there is a bad sign for Democrats -- I was thinking it's a bad sign for the reputation of prediction markets. There's something about Hillary that drives ~10% of bettors insane, which the only reason her name is even at 2c up there.
Assuming I were 100% confident that it goes to zero, I'd have to spend $980 right now to make $1000 when the contract resolves in 2.5 years, which turns out to be less than a 1% return.
Scott was talking about trying to bootstrap returns from far future events, but there's only so fast you can bid down something with this small of a return.
It's not about making money, it's about prediction markets as a means of finding truth, which is what Scott is advocating for in this series. Hillary has nowhere close to a 2% chance of getting the 2024 nomination. She had nowhere close to a 10% chance of being arrested and jailed. She had nowhere close to a 10% chance of getting the 2020 nomination (nor did Yang, for that matter). But those were the chances that the PredictIt markets settled on. You can't expect people to trust odds set by prediction markets when those markets are frequently distorted by cultists and conspiracy theorists.
I also agree with that, but the prediction markets would have found some other names to say instead if there were realistic candidates. That speaks to the idea that they don't think Biden will be the candidate, but really can't think of someone who makes sense that is actually plausible.
I'm curious to see if the Reddit prediction feature works. Getting prediction tokens seems like it would be good for clout in a subreddit, but the strength of that incentive probably varies a lot across Reddit. Too bad PredictIt and PolyMarket don't have subreddits. If Reddit can beat a PredictIt market on a question that matters, that would be really surprising, because seeking clout on a subreddit would be a better incentive than money to reveal hidden information and make accurate predictions
>I’ve heard a lot of stuff about the prosecutor really bungling this one, but mostly from conservatives who I would have expected to hate the prosecutor anyway, so it’s good to get objective confirmation that yeah, this isn’t going anywhere.
I'm not saying the prosecutor is doing a great job, but the main disconnect here is between "he did something bad" vs "he did something illegal." The law sets out specific criteria about what counts as self defence, and I think Rittenhouse both was irresponsible and definitely hits the bar for self-defence here.
Think of it as similar to QI things, where people are shocked cops aren't convicted for doing bad things. We wrote laws/made SCOTUS decisions saying you can do *extremely* bad things without it being a crime. I don't think the prosecutor screwed up, I think he's unconvictable on the merits of the law.
I think because it was so controversial and trying the case means people can see the adversarial process play out and hypothetically have peace with the outcome, vs it being the prosecutor’s private call.
One possibility is that there's room between "no reasonable chance of conviction" and "conviction at trial is likely". If the prosecutor judged something like a 20% chance of conviction at trial, it might make sense to proceed anyway. If the prosecutor believed Rittenhouse deserved to be punished (either because he's likely in the prosecutor's judgement to be factually and legally guilty even if not necessarily to a reasonable doubt threshold, or because the prosecutor is assessing dessert based on abstract notions of justice or cynical judgement of the needs of public order rather than a faithful adherence to the letter of the law), a 20% chance of conviction is better than a 0% chance if you don't bring charges.
Moreover, the threat of going to trial when there's some chance of conviction but not a great chance is how prosecutors get leverage for the plea bargaining process by which they secure most of their convictions, giving a cynical reason to prosecute a trial you're likely but not certain to lose. It's like being the Dread Pirate Roberts: once word gets out you've gone soft, it's nothing but work work work all the time.
Both of the above are compounded by the possibility that the evidence looked better to the prosecutor before trial. Key testimony from prosecution witnesses seems to have come out more favorable to the defense at trial than the prosecution was expecting, and also the judge made some rulings on evidence admissibility that were likely more favorable to the defense than the prosecution was expecting.
Even if the prosecutor did believe he had no realistic chance at conviction, there are plausible cynical reasons for going to trial anyway. Both for public order considerations and for his own reputation and career.
Going to trial could be an exercise in failure theater. If the prosecutor dropped charges before trial, it's look to members of the public who believe Rittenhouse to be guilty like the prosecutor (and the system he's part of) was complicit in letting a murderer go free. This make the prosecutor himself a villain, further undermines shaky confidence in the system, and risks triggering more violent protests in response to the perceived injustice. Taking the charges to trial gives a chance of shifting perceived blame from the individual prosecutor and the legal system as a whole to the judge and the jury.
There's also a potential angle of "the process is the punishment". Facing murder charges and the possibility of life in prison for a year or so isn't nothing, even if it's enormously preferable to actually being convicted. If the prosecutor believed Rittenhouse deserved punishment, or the public order required Rittenhouse's behavior to be disincentivized, then putting him through the rigours of trial is one way to do that.
Interesting. My impression was that they knew they couldn't win on it (due to a lack of any effort at enforcement), and they brought it just to imply that he was there illegally and add to the perception of his recklessness / lawlessness; and actually arguing the charge would have undermined that perception.
Same with the gun charge, finally thrown out today. The statute in question doesn't apply to the gun he was carrying (not a "short barreled rifle"), nor to him as a 17 year old, I think. (I believe it has an exemption for 16 and 17 year olds.) But the prosecutors brought both these inapplicable charges to better hand-wave at "illegal gun, illegal presence, reckless conduct". Interested if you have a source contradicting that.
The gun statue is confusingly worded, and both sides think that it obviously applies or doesn't apply to Kyle. From what I've heard, the judge threw it out precisely because it was so confusing.
As for the curfew charge, the impression I got is that they'd have had a slam dunk if they bothered, but I'm not a lawyer, so I have no idea how those things actually work.
Isn't there a problem with the curfew charge, that there were literally hundreds of people actively breaking the curfew, including the people who Kyle shot/was being chased by?
Not very sympathetic to go after Kyle only on a curfew charge. I also find myself wondering if the lack of prosecution of the guy who got shot in the arm for having an illegal weapon there may influence the jury?
If the gun is a rifle or shotgun the charge doesn't apply unless it's short barreled or the individual is younger than 16 and in violation of 29.593. And since he's over 16 and the gun isn't short barreled it seems like he can't be in violation of the law.
The language is "not in compliance with 29.593", not "*in violation* of 29.593". Those aren't necessarily synonymous, because it isn't clear whether one can be "in compliance" with a law that isn't relevant to the issue at hand. Is a pilot flying an airplane over Milwaukee "in compliance" with Green Bay municipal traffic regulations? If you think the answer is obvious, then either you know more about Wisconsin law than e.g. a Wisconsin judge, or you don't know as much as he does.
The legal arguments I've seen that he was in violation of the law weren't arguing the case you made here but a more complex case around legislative intent. Basically that the law as stated says he's not guilty but that was a mistake of the legislature.
Having watched an unhealthy amount of the trial footage, I agree with your overall assessment, but I think it is still soft on the prosecutors.
In particular, ADA Binger's antics, including trying to bring in excluded evidence and questioning Rittenhouse about remaining silent pre-trial (a big 5th amendment no-no), led to some dramatic moments last Thursday. It got to the point where Binger gave his rationale for his actions and said they were in good faith, and the judge looked him dead on and said (inexact order of phrases), "I don't believe you... you're an experienced prosecutor. I don't believe that was in good faith. I'll take the motion [mistrial with prejudice] under advisement. And I don't believe you." I think it's clear that he's hurt his own case with his shenanigans, quite possibly to the point of invalidating a guilty verdict if one is delivered.
That said, I've seen people claiming that the prosecutors should receive professional discipline, or be tried for misconduct, and stuff like that, which is ludicrous. As the saying goes, they can't pound the facts, they can't pound the law, so they're pounding the table. There's nothing punishable or even remarkable about it.
There seems to be a combination of "DA realised this thing was toxic no matter what way it goes, so he handed it off to Binger" and "Binger is ambitious, ran for DA in another county, and wanted this as big case to make his name" at play.
> I don't think the prosecutor screwed up, I think he's unconvictable on the merits of the law.
Then the prosecutor did screw up for laying those charges the begin with. They try really hard to only go to trial when there's a chance in hell of winning.
Yeah, there have been some “bungles”, but when the most significant “error” is allowing your alleged victim to described what happened to them, you probably didn’t have a great case to begin with.
It did get bad enough that some people (including the judge!) implied the prosecutor may have been angling for a mistrial. But that was because the prosecutor was trying to slip in some disallowed lines of argument (again, less an error than an admission of a weak case).
Another in the "experts get things repeatedly wrong" category: all the big investment banks have teams of analysts who try to predict how much profit public companies will make in the future. It is well known that these forecasts are consistently wrong -- specifically they are upward biased. Interestingly this bias persists despite there being a significant financial interest in getting the numbers right, since funds often rely on these figures when making stock picks.
One theory for why this is so is that the analysts intentionally publish bullish forecasts to flatter company management, which helps them e.g. get meetings with the executive team.
I doubt anybody puts much faith in them, given that there are already prediction markets for how well a publically traded company will do (aka markets)
>and decided the best paper out there was Brauner et al
I'd question whether these kinds of questions can be measured by science (or superforecasters) at all, but I'm a bit confused that "closely correlated estimates" is supposed to be impressive. The point of the estimates isn't to be closely correlated, but to take approximately the same values, no?
The "select areas" are fairly widespread, but focused on regions that don't already have great internet, which explains why California isn't a priority. Even the best Satellite internet will still have fairly high latency, which does limit its usefulness if you have access to fibreoptic lines.
Addendum, I actually went and looked it up, and the current satellites cover the USA, Europe, Australia, New Zealand, and Chile, as best I can tell. Not quite sure what Chile is doing in that list tbh, the rest are all wealthy countries. Also not sure why bay area isn't available yet if there are satellites covering it...
"Their solution is to make teams very big and full of smart people, so that it’s unlikely other teams will miss something."
Because I am an incurable smartarse, I can't help feeling about prediction markets as per "The Napoleon of Notting Hill":
“The human race, to which so many of my readers belong, has been playing at children's games from the beginning, and will probably do it till the end, which is a nuisance for the few people who grow up. And one of the games to which it is most attached is called "Keep to-morrow dark," and which is also named (by the rustics in Shropshire, I have no doubt) "Cheat the Prophet." The players listen very carefully and respectfully to all that the clever men have to say about what is to happen in the next generation. The players then wait until all the clever men are dead, and bury them nicely. They then go and do something else. That is all. For a race of simple tastes, however, it is great fun.”
RE: the media all claiming the Virginia race was too close to call - well, the final result was Youngkin with 50.6% of the vote and McAuliffe with 48.6%, so it was pretty close. But my cynical interpretation was "These are liberal to left media news organisations, naturally they will be biased towards 'No, it cannot be that the uncouth Republican will triumph in Blue Virginia, so we will not give the oxygen of publicity to these threats to American democracy".
As has been pointed out, it's more that all their incentives point towards not calling any race until they have no choice. A) viewers want to watch a race, not a victory lap. B) being right gets you nothing, but calling an election wrong gets you excoriated
I'm going to push back on Erik Hoel's prediction "#17. People and culture will become boring." TLDR: Because it will be possible for a single individual to create entire movies or series in much the same way a single person can create a book right now, so we'll have an explosion of new ideas in media.
He argues:
>The panopticon of social media and state control will lead to cultural stagnation. We already see early hints of this. Consider the remakes of older movies: 2050 will be a stew of remakes of remakes, and familiar and boring intellectual property (like Star Wars) will be king. Creativity vivacity will suffer, especially in the arts and humanities. The coming half-century will be a great one for innovations in finance, engineering, space travel, and artificial intelligence. It will be a terrible one for the arts and basic scientific advancements (like a new physics), for such advancements require iconoclastic and creative lone individuals. This prediction is already augured by judging the 2000-2020 creative period overall in areas like art, music, literature, film, and scientific discoveries, and finding it severely lacking compared to, say, 1950-1970.
I think the reason we've seen less creativity is for economic reasons, not cultural ones. It's just so much money and work to make (and market) a good movie that studios play it safe and mostly stick to works they know will be an economic success.
But by 2050, AI will have made it so a single hard-working individual will be able to create an entire movie by themselves. The computer will render the scenes described and the 'author' will be able to direct changes until their creative vision is fulfilled.
This will make movies like youtube circa 2008 when there's all kinds of new opportunities for creative people and tiny little indie director/authors will go viral. Eventually the movie media will look more like our print media. There will be well established author/directors making a new movie every couple of years in far more genres than exist in current cinema. And it will be much easier to be profitable with creative niche projects so more people will be willing to try them or create them for art's sake or as a hobby. If you only need 20,000 people to purchase tickets to a movie to make it profitable, it opens up the room considerably for more creative media.
At the same time, VR will open up a whole new space for new experiences. It's not a coincidence that early cinema was an age of creativity as it takes a lot of experimentation to figure out what works. VR is going to go through the same process as it slowly becomes a major source of our media.
And so eventually, we'll consider the years from around 2040-2060 to be a golden age of creativity.
You don't need the average guy to be creative. You only need the top people. If it's possible for the top 1% of creative people to make a living doing solo movie projects and only 1% of those people decide to make movies for a career that's ~50,000 author/directors. And potentially 10s of thousands of 'indie' movies coming out per year. And you'd only need the average American consuming a few of those movies (as paying customers) to make it profitable, which seems plausible.
Early YouTube just had a huge variety of creators all testing out the platform. So there was lots of variation and creativity.
I’m very dubious about the top 1% of people creative enough to produce worthwhile art - that’s close to 3 million people in the US. The number of great directors alive is closer to a dozen.
And are we in a golden age of art right now because of new technology? What art has been solely produced by YouTube?
And since mass literacy everybody has has the ability to become writers, the result hasn’t been an explosion of great writers relative to population, but the reverse.
Freddie Wong and Corridor Digital would be my go-to examples of indie filmmakers on Youtube. They've made some very impressive shorts with a very small crew and budget.
(Corridor has a bunch of behind the scenes videos, and probably half of them mention how a traditional production would have done [expensive thing] to get a certain look, but they've substituted some VFX magic instead.)
"VFX magic" often isn't, unless it's done sparingly and very, very well.
And while someone might be willing to overlook flaws in an amateur, small-scale project, when that is now allegedly a professional production vying for paid market, they will expect more and be less tolerant of "you could see this was greenscreen!"
I generally can't notice on a casual view, I would have to pause the video and spend a couple minutes thinking "should the shadows *really* be falling in that direction?" to figure out what's been composited. And they have gotten into the paid market (not just ad revenue, like, actually paid to produce a show), which is a more objective measure of competitiveness.
But my point was more about the budget - the OP's prediction was that improvements in CGI software and AI would allow a single author to create a whole video by their lonesome, and Corridor's shorts are what I'd call the first steps towards that vision. It's nine guys instead of one, and the cameras and gear aren't exactly cheap, but it's getting better all the time.
Yeah, but if I go onto Apple Store or Netflix or wherever, there are a *ton* of movies I heard nothing about, had no idea were even produced, and I have no interest in watching. You're forecasting an explosion in even more amateur movies competing for people's attention.
How many blogs are there? How many Twitter accounts? How many Instagram influencers competing for money and attention? A few will be successful but the majority will be scrabbling for pennies (relatively speaking). It may or may not be a burst of creativity, but getting the attention of even 20,000 people for your one project is going to be exponentially more difficult due to those tens of thousands of indie movies all wanting their eyeballs. You're more likely to fragment the market, with maybe 1,000 people paying to stream "My Brilliant Take On Noir Detectives In Space" over the Internet, rather than 20,000 willing to pay for movie tickets in a traditional theatre and going out to watch that movie. And the prices will have to drop correspondingly, who wants to pay you $10 when they can watch all that free content others are churning out?
What you describe has already happened for writing. Anyone can write a novel, self-publish at no cost, and have it available to everyone on Amazon. Has the quality of fiction gone sharply up as a result?
It's true that writing a novel was always something that an individual could do but he also had to get a publisher to publish it, which was a pretty severe filter. So not exactly equivalent to your case, but in the same direction.
Good point, but it's not the ability to publish that I'm thinking of so much as the ability to create in the first place. One person has always been able to create a book by themselves and I think in ~20 years one person will also be able to create a film by themselves. They'll probably still be gatekeepers for those films like there's gatekeepers to books but there's currently a much wider range of books than movies and I expect that gap to shrink.
Plus, there are social talents an author-director of movies has to have which would be obsolete for single creators, like a medieval bard or even Shakespeare with his theatre business compared to a modern novelist like Dostojevski. The productive talent pool may well get bigger.
A lot of my favourite books recently have been self published, so it feels like it to me. I greatly enjoyed The Martian, Worm, Mother of Learning, and Harry Potter and the Methods of Rationality. Maybe it's because they're all a bit different than what typically gets published.
I think the filter you talk about is in the wrong place if you intended it as a counterpoint. Many lone authors create books and then publishers (ideally) choose the best ones to publish. Assuming publishers choose well, no quality is lost.
With movies, most people with a great idea will never even create a movie to begin with.
Hell, even among published books many successful works were famously rejected by multiple publishers before finding a home, eg. Harry Potter, and there's a hell of a lot of published trash, so I think traditional publishers are a terrible filter!
Editors, on the other hand, are invaluable... AI shouldn't be too far off being able to edit for spelling and grammar, surely, even if editing for plot and tone will still require a human for a while?
Yes, but as those AI will all be controlled by giant woke corporations, anyone trying to make anything problematic will likely have their license to use AI stripped from them.
I don't see any obvious way for Google to stop a non-SJ Google-clone from catering to the people Google rejects. Their AI might be a few years behind and 1/10 the scale, but they'll have it.
You need actual law (or terrorism; violence, in any case) to push things beyond "separate-and-not-equal" (the current state) into outright "no access to services", because the class of SJ excommunicees is large enough to form its own economy and can't be dissuaded by status incentives.
There's an excellent example of this already. AI Dungeon is a text predictive choose-your-own-adventure game that uses GPT-3. Recently, OpenAI (the owners of GPT-3) forced them to implement a heavily restrictive content filters, to the chagrin of most users. Disgruntled former fans worked to build a competitor called "NovelAI" using open-source GPT clones, EleutherAI's GPT-Neo and GPT-J, which are 2.7B and 6B parameter models, compared toGPT-3's largest model is 175B parameters.
Despite having a model 1/30th of the size of AI Dungeon's, NovelAI is superior in many respects. While its raw text prediction is slightly inferior (but not by that much -- GPT-like models achieve diminishing returns fairly quickly and EleutherAI's models roundly outperform similarly-sized OpenAI models in most tasks due to far better training data), NovelAI compensates with a ridiculous amount of quality of life and customization features that, in the hands of an experienced user, can give superior results in general to AI Dungeon despite an objectively weaker AI.
So you're forecasting no more live-action actors? It'll all be CGI and animated models?
Maybe we will have the equivalent of self-publishing today, all the ebooks on Amazon etc. But an awful lot of those are absolutely crappy content, I am willing to bet that most of the "single individual makes an entire movie" will be crappy content as well (I don't just mean the writing, I mean layout and design and typos and lack of editing etc.)
So I think Hollywood will still have an advantage, even there: if everyone can make a movie without actors, the established industry can do it faster, better and cheaper, with the brand name recognition: "Do I want to watch a new movie from Paramount or from J. Random Nobody?"
I think first it will be cartoon/animation that it will be possible for one person to create with the help of software. And eventually the rendering will be indistinguishable from live actors. But probably at no time, and definitely not in the next 50 years, will there be no more live actors. It will be a both thing.
They'll still probably be big Hollywood movies and studio TV shows produced by big teams of people and actors. But they'll also be a huge amount of independent mostly solo created content (in a similar way to books are now). The best of that will get picked up by things like Netflix and distributed widely, which will infuse the wider-culture to some creative diversity.
Probably the biggest market though will be 'better remakes than Hollywood.' So just like fan fiction is big now, they'll be tons of different non-canon sequel trilogies of Star Wars or different endings to Game of Thrones, or additional stories in he Harry Potter universe.
Right, Hollywood will always have more polish. But do I want to watch the 1000th polished Marvel movie, or a somewhat flawed indie darling that is actually original? The explosion of indie web novels makes word of mouth and reviews much, much more important, but there are true gems out there to find, and I expect movies will end up in the same way, if not more so! (more so because conventionally published novels show more originality IMO than mainstream movies - precisely because of the different level of economic risk involved, I expect.)
I'm going by Sturgeon's Law: 90% of anything is crap.
Yes, there will be great indie movies. But you'll have to wade through an ocean of crap to find the few jewels.
Hollywood will give you a reliable product. It may not be original but you know what you'll get.
I'm not denying there would be great original 'made this in my bedroom' stuff, I'm saying that if every Tom, Dick and Harry is doing it, you're going to find it very, very hard to go through all that content and find the good stuff. It'll be more likely the really good ones will eventually be picked up by Hollywood - or Netflix, Amazon, HBO, etc. to produce content that can be easily mass-marketed.
You have 1 million indie shorts on your Youtube algorithm curated recommendations list. Are you going to wade through all 1 million or are you going to pick "I really like Tom Thompson's stuff" and stick with that niche of "stuff like what Tom Thompson makes"? Because I think that is how it will end up.
"I guess at the limit this is just banning insider trading, which is supposed to be good."
I think most economist think banning insider trading is bad for getting the market to reflect the best guess of where the price should be because it is leaving out a piece of true information held by the insider. So it does make the prediction market work worse.
I think the argument against insider trading is more of a principal agent problem. Should the insider be using his privileged position to be making money of the stock holders who he works for.
The two are not mutually exclusive. "Not wanting to look foolish by calling a race too early" seems a very plausible media bias.
Likewise, I expect most news outlets to be biased in favor of their stories getting more click and views. "Too close to call! Stay tuned for continuous updates!", if it seems plausible to casual viewers, would probably get better ratings than "Welp, looks like the Republicans have this one in the bag."
I'll note that if I were pro-Democrat, and I believed the standard model that Democrat turnout is more mercurial than Republican, I'd report everything remotely marginal as "too close to call" to get out the vote.
It's not a more likely explanation. Too close to call is very common terminology that's been used by media election desks for at least the last few election cycles, independent of which party is leading.
"even if you double your money, 100% return over eighty years is a bad investment."
Honestly, this is a problem over even two years--a 10%ish opportunity cost is a lot.
But can't this be resolved by purchasing conditional shares in some portfolio where the shares are traded freely? Like an SP500 index conditional share?
So I purchase a conditional share in a "China GDP 2100 vs. US GDP 2100" SP500 index fund. The value of my share increases in lockstep with the increase in the fund, which over the course of 100 years is a fine return. Unconditional fund shares can also be bought/sold as usual. In 2101 the conditional shares will turn into real shares, or into nothing, depending on whether the condition was met or not.
This seems to solve the liquidity and returns issues both.
(Tax loss harvesting from losers might skew some numbers temporarily, but that's even a good thing in that it provides non-arbitrage liquidity)
But I'm obviously not the first person to think of this, so...why isn't it standard?
The media has a well known incentive to always describe races as closer than they actually are so that people will keep breathlessly reading updates. The same thing happened in the 2012 presidential election.
This is everyone(-ish) said about Nate Silver claiming Trump had as high as a 30% chance the weekend before the 2016 election: "He's just doing it for the clicks!"
The biggest weakness in Forecasting today is the unknown unknown, meaning, getting the most relevant questions forecasted in the first place. For instance, it's more relevant to the world to forecast the probability a pandemic will occur in the next 3 years then to predict the level of active Covid cases in 3 months, given a known Covid pandemic. Also would have been extremely useful if forecasters had foreseen the supply bottlenecks 6 months in advance (because then they would have been less likely to have happened, given that forecasting markets had high enough visibility. No, this wouldn't have been a paradox: the odds could have risen and declined over time as the forecasting market spilled information into the market market.
A pandemic was a black swan, fair enough. But the drop in industrial production a year ago should have given more than a few people the bright idea that: when this thing is over, demand will come roaring back and when it does inventories will be low... and there will be a crush at the ports. That's wasn't a black swan; that was foreseeable. I didn't see it, but I'm sure others did. Perhaps they kept it quietly to themselves and put their money into crude oil futures...
Yeah the trick with unexpected events is that no one in their right minds really bets on them. Hedging is different, of course. I agree about the adaptation phase. When things started to get messy, one had to be more imaginative and less risk-averse than before to forecast stuff.
"Prediction markets reached near-certainty about the winner while traditional media was still talking about how un-call-ably close it was. Apparently having hundreds of people all incentivized to give precise probability estimates very slightly earlier than the next guy, works better than having a few journalists who are scared people will make fun of them if they jump the gun."
Compare this to how PredictIt wasn't willing to call the 2020 election for Biden until long, long after it was obvious that Biden had won. I think a better model might just be that PredictIt and some similar markets tend to err in the direction of Republican candidates for contingent historical reasons.
I wonder how much that had to do with 2016 and every media pundit and pollster serenely certain Hillary would win because look at her leads in the polls, look how she is forecast to win Florida etc., and then they got a massive pie in the face on election night.
So they erred in the other direction of "well since not *every* single vote has been counted, let's not announce anyone as the definite winner".
At least one prediction market had Hillary at 89% the day before the election, too, which would have been a chance for Scott to make some of his money back on bets against extreme prediction market certainty.
Amplifying small changes can make the smart people lose all their money faster when markets temporarily move against them.
Also, depending on the exact mechanism for resolving each 5-year market, which you don't really explain, the resolution could be manipulated. If there is a lot of open interest on the expiring contract relative to the volume on the new contract, people can profitably manipulate the price of the new contract to manipulate the settlement of the old one.
> Might prediction markets be lurching to a false certainty too soon? I used to think this, I tried betting on that thesis a bunch of times, and I always lost money. I guess realistically we can’t know for sure that they’re not overconfident until we’ve tested their 98% probability fifty times, but I nominate someone else to lose their money for the remaining 40-something experiments.
At the beginning of November last year, Biden was predicted to win probably but low-confidence. On the night of the third they were medium-high confidence (but less than 90%) that Trump would win. By the time they got to 95% on any question I think they were usually right, but they definitely were prematurely confident.
Sanders is vastly overrated. Unless people are taking early cheap bets to score off his small but extant cult base. Warren is immensely overrated as well. They should both be at 1, as should Clinton. AOC is slightly overrated. Buttigieg is underrated. Harris overrated and Biden underrated. The real race is in dark horses who haven't even talked about it yet. Also the PredictIt markets are notoriously poorly moderated/modified.
Regarding Hoel's prediction #14, here's my honest counterprediction:
30% nuclear war by 2030.
I'm not about to bet in a prediction market on this, and I indeed predict that prediction markets will predict against it. This is because all predictions on prediction markets are conditional on the prediction market itself continuing to exist, which is highly correlated with the outcome of some questions. But it's why I don't live in a large city any more.
On street votes, it's worth clarifying that the even most ardent advocates do not think that ‘most’ people will vote for them. Even the Strong Suburbs report linked to in the text above does not estimate an overall take-up rate of more than about 3%. The estimated price elasticities implied that that 3% would exhaust the economic potential for profitable densification. The potential to add more homes well in low-density suburbs is so great that that low take-up can mean the production of a considerably higher volume of new homes than happens under many current zoning systems. Many suburban lots are easily capable of producing 5x (sometimes 10x) the current amount of housing on them, even without going above three or four floors.
There may be an analogy with cryonics here. I think one survey showed that the people who had signed up for cryonics assigned an even lower probability to cryonics working than did the people who had not signed up. The former group had just looked at the numbers and concluded that the potential upside was greater than the downside.
I guess I would be concerned that locals would be even more anti-development than central planners because they're the ones who have to deal with the noise and traffic. It sounds like you're saying YIMBYs also expect this but think that increasing variance (by coincidence a few blocks will be mostly YIMBY or neutral) is worth it?
Yes, it's partly about variance, in two ways. (1) Some streets will be much less risk averse than others, as you say. (2) Even if homeowners are on average more anti-development than planners, then we get a double option on each potential development; either it can be allowed by the planners or by a small group of homeowners. The proposed legislation doesn't let homeowners block development that the planners allow. So it's strictly superior in terms of housing supply. And many of these suburbs produce almost no new housing at all at present, so it would be difficult to reduce that even if there were unforeseen consequences and different legislation in future.
But it is primarily driven by the details of the process, on two principal dimensions.
Number: many individual homeowners are willing to do more with their lots, including selling for redevelopment (at a potential profit of over a million dollars per lot in some areas). If they weren't, there would be less demand for zoning rules to protect other homeowners. But the incentives are far more diluted and harder to obtain when the decision requires a large number of people. It would be theoretically possible to rezone the whole of Palo Alto and make every homeowner much better off financially because of the enormous latent demand to live there, but they're likely to see it as risky and it would be incredibly difficult to get so many people to agree. With ten or twenty homeowners, the incentives are vastly stronger – nearly as strong as for a single homeowner – and the costs of persuading them much lower.
Mechanism: one-shot direct democracy as we are proposing inherently involves fewer veto points than layers of committee-style structures. There is much more likelihood of blame avoidance where there are multiple primaries and other selectorate mechanisms operating to make officials and politicians more risk averse. Persuading committees of people who have to get re-elected to take real and possibly controversial action is hard. And many zoning systems have accreted layer after layer of veto points in the process, with the intent to reduce controversial change.
We have good evidence from Tel Aviv's TAMA 38 process and from Seoul that people do vote for redevelopment through direct democracy, even with the 75% majority as required in Seoul (e.g. https://capx.co/seoul-searching-does-the-korean-capital-have-the-solution-to-the-housing-crisis/). And we know from Nolan Gray's work on Houston that many streets and blocks did not take up their right to opt out of the minimum lot amendments that allowed more housing there.
So the downsides seem minimal and if we achieved even a fraction of what was done in Tel Aviv or Seoul, it could substantially improve housing supply.
The American Planning Association (many of whose members are enormously pro-housing and are presumably familiar with their local political obstacles) thought it was interesting enough to invite us to write an article recommending people try it. https://www.planning.org/publications/document/9219200/
Erik Hoel's prediction #1 (that there will be a Martian colony in 2050) is not only embarrassingly ridiculous, it doesn't even follow his dictum of -
"If you want to predict the future accurately, you should be an incrementalist.."
An incrementalist would look at what proportion of the "getting a human being back from Mars, alive" problem has been solved in the 50 years since we put a man on the moon. And the answer is just a few %. The task, with current technology, is not remotely possible - which is why NASA talks (or fantasizes) about nuclear propulsion in space or setting up a mining industry (remotely, with robots!) on Mars to provide the fuel to power a return journey.
The irony is that after criticizing other smart people for "..imagining a sci-fi technology that doesn't exist" he then goes ahead in his very first prediction and does the same. He simply "imagines" that currently completely unsolvable problems will suddenly become easy-peasy such that there will be "a city on the red planet".
No there won't! There won't be anyone on the red planet in 2050!
I'd add that I think what Elon Musk has achieved with SpaceX over the last 20 years is nothing short of incredible. But he hasn't even begun to solve the really big conundrums that make currently getting to Mars and back (for humans) completely impossible.
Musk really hasn't broken any new ground that wasn't completed by the 1960s. He's done it cheaper, and he's done it in a way that's easier to replicate. At this point, his company is acting more like an engineer than a scientist (refining rather than inventing).
If, and that's a pretty big if, he keeps going at the current rate, there's a good chance he'll progress beyond what humans have done in the past. So far, he's mostly just treading old ground. He hasn't even done some of the major things that were done 50 years ago, like going to the moon. Or more recent items like going to Mars, or sending a probe out of the solar system.
Funnily enough I think the one major thing he has done is in the direction of making going to Mars more feasible - he almost alone in thinking that heavy lift rockets need to be re-usable. And I think the progress SpaceX has made in that direction has been truly ground-breaking. Sure, there are a dozen more major hurdles, but it more progress than anyone else has made in the last 50 years..
Colonization implies moving there permanently; living on Mars for decades doesn't necessarily require the ability to return but it would be uncharitable to describe such a trip as "Die when you get there."
Sure, fair enough. However the first trip is absolutely only going to be made if it includes a return. Which is another reason it isn't going to happen anytime soon. And for sure not before 2050.
I mean, no-one was really trying to advance manned space flight between ~1970 and ~2010, so it's not that surprising that there was no progress and some tech even moved backwards. Given that Musk is willing to throw billions at getting to Mars, it's not that implausible that he might succeed - historically, the Apollo project showed how fast tech can advance with money thrown at it, and the lack of anything since then is squarely due to a lack of investment.
I'll grant you that the achievements of the Apollo project always give pause to Mars scepticism. For me, the progress From Kitty Hawk to Apollo 11 is even more arresting.
However, just throwing money at problems doesn't guarantee solving them, or solving them within the next few centuries. Nuclear propulsion in space may indeed work, but then again it may not, or at least it may take hundreds of years and trillions of dollars of investment. If not nuclear it'll be fuel produced on Mars which is Musks preferred solution, and the evidence for the feasibility of that is almost non-existent.
To take the other major impediment to getting to Mars, the record for the length of time a human being has survived in deep space - in the radiation 'death zone' is 12 days. And though I take your point that nobody has been trying to extend this for the last 50 years, we still don't have the faintest idea how people will survive 500 days of severe cosmic radiation let alone solar flares. We might find a way, but it might take (again..) centuries.
I wish Musk and NASA and anyone else who tries to get a human being to Mars and back all the luck in the world, but I'm not holding my breath.
My prediction is that there will be people attempting to colonize Mars by 2050, and most (and perhaps all) will die either there or in transit. And they will be lauded as heroes.
By comparison, consider the bottom of (Earth's) ocean. There is considerable mineral wealth there waiting to be utilized, it's orders of magnitudes closer, and orders of magnitudes less hostile to human life, but I don't hear about efforts to build colonies there. Why? I think it's all about glamour.
Scott has written in 2015 in a short fiction story about (among other things) a person who could see one month into the future:
> You resolve, on the first day of every month, to write down what you see exactly a month ahead of you. But what you will see a month ahead of you is the piece of paper on which you have written down what you see a month ahead of that. In this manner, you can relay messages back to yourself from arbitrarily far into the future – at least up until your own death.
So he has already invented iterative prediction 6 years ago!
P.S. The story is called "...and I show you how deep the rabbit hole goes"
“Too close to call” means what, < 99% confidence? Because no media outlet wants to retract a call, ever, and there’s no penalty against being late with a call.
I think a lot of prediction market design problems are already solved in financial markets. Suppose you have a non-dividend paying stcok (say, Google's parent Alphabet Inc.) The fundamental value of this stock is, theoretically, a function of very distant future cashflows, it is the present value of dividends that might be paid in 20-100 years from now, when that company will stop its crazy growth and will start distributing cash to shareholders. This long (and uncertain) horizon does not prevent a liquid market in Alphabet's stock. Indeed, right now this market is able to assign a very tight range of values to this present value, $2978.00/$2979.84.
How does the financial market handle the very real impediments to prediction markets that you describe?
1) Removal of intermediary credit risk. You can be right about your prediction but the probability that a centralised provider of a prediction market will fail grows quickly with time horizon. I would not bet more than even odds that a single current centralised betting market provider will survive until 2100. So one can be right about US vs Chinese GDP in 2010 and still lose 100% on this bet when market for the chained bet #6 fails in 30 years time. This problem can be resolved by DeFI ("can be", not "has been" as I am not sure we can be that certain about existing protocols.)
2) Diverse participants time horizons. For a market to provide a liquid tight price one needs to have participants with diverse time horizons. Market for Alphabet stock is provided by people who intend to hold this stock for less than a second. Plenty of traders hold this stock for less than a day. This should not be a problem in most prediction markets.
3) Availability of leverage. 100% return over 80 years is, indeed, a paltry reward for a correct long-term prediction. This does not create a problem for equity markets for two reasons - the mixed horizons mentioned above and availability of leverage. The money invested in a liquid contract is not frozen until contract maturity. One should be able to borrow against market value of this contract at pretty low interest rate. This is not something available today, but there is no reason for this to be unsolvable by DeFi.
4) Fragmented markets are less efficient. Equity market is much more transparent and efficient than bond and option markets. The main reason for that is that for each corporate there typically only one stock but dozens or hundreds of different bonds and thousands of options traded, so liquidity is spread much thinner. Chained iterated market, especially one with bins would fail on this count quite badly.
Thus, I do not see why a much simpler solution to a prediction market for US/China GDP in 2100 would not work better than an iterated market. On a trustworthy DeFi exchange, create a contract expiring in 2100 and settling at 0/100 depending on the binary outcome. If this contract becomes liquid enough, it will attract traders with diverse time horizons. It will be also relatively stable and it will be possible to borrow against it, so the limited return will not be major problem. To enable higher leverage, somebody might introduce options on this contract, may be even binary range options you are describing, but the contract can be perfectly liquid without them.
#1 requires that the oracles that the defi contract points to still exist and are still trustworthy. That's just as much of a longshot as PredictIt still existing in 2100.
#4 is a good argument against chained prediction markets
Ah ha! I wondered where all those new subscribers were coming from! Really and truly appreciate that shout-out Scott, as I'm a longtime fan. It's also insanely impactful to a smaller substack like mine - I think around a 20% increase in total subscribers in the ~24 hours since this post went up
Is this sarcasm? Or maybe you mean only in abstract theory?
> My only concern is that it doesn’t have the same sort of hits-you-between-the-eyes obviously-there’s-no-way-to-bias-this quality that prediction markets do.
I always thought prediction markets had the opposite quality: that they're super easy to bias. I admit I haven't kept up with the research on this, but prediction markets of real-world events lack of the property that people's predictions have to converge on the actual answer, because:
* people aren't really financially constrained by predicting wrongly, since they can just get money elsewhere and aren't going to go broke on bad predictions, so over time bad predictors will stay in the game
* adding more information (or getting closer to an event) doesn't necessarily make people better at predicting what the outcome will be, due to people's tendency to just.. be wrong, misled, or biased (not a property of some types of predictions, like things that have some random distribution, or where polling or something provides an increasingly accurate prediction of the outcome).
(arguably investment funds have the first property, for comparison, because they deal with quantities of money at which bankruptcy is a real risk.. but the real-life dynamic that old funds go broke and new ones are founded might mean that overall the market is full of funds that won't not survive actual market fluctuations over time; they've not had time to be 'weeded out' yet.)
> If there are a lot of low-quality forecasters in the tournament, then since high-quality forecasters will accurately predict that low-quality forecasters will give a low-quality answer, everyone will converge on the low-quality answer.
> They also admit this incentivizes teams to ignore “secret knowledge” that they have but which they expect other teams won’t. Their solution is to make teams very big and full of smart people, so that it’s unlikely other teams will miss something. I guess at the limit this is just banning insider trading, which is supposed to be good.
Isn't this what Bayesian Truth Serum is supposed to solve?
In the Erik Hoel link, the graph about single parenthood seems to be showing a different statistic than he says it shows (percentage of single adults who have dependent children versus percentage of children raised in single-parent homes; although these are correlated in practice, it would be theoretically possible for one of them to be close to 100% while the other was close to 0%; and both of these are different again from the percentage of children raised in single-parent homes, which the post also mentions). Also, the post refers to a percentage figure that isn't shown on the graph ("In Denmark, for instance, the total percent of single-parent households in 2019 was around ~33%"; even given the previous point and replacing "percent of single-parent households" with "percent of single adults who are parents", the graph shows about 20% of single women and about 8% of single men in 2019 were parents, so about 14% of single adults overall, so I don't know where 33% comes from).
If I saw this in a random place on the internet I'd just assume the author didn't understand what they were talking about (the first issue is similar to the widespread confusion between P(positive test|disease) and P(disease|positive test), and the ~33% could possibly have come about by thinking that the overall percentage comes from *adding* the male and female percentages, which would give 28%, possibly combined with some misreading of the chart axes to get it from 28% to 33%).
But since this was linked from ACX and recommended by Scott, and since none of the other commenters either here or there (who are clearly highly numerate and also nitpicky) have picked up on it, my prior for it being my misunderstanding is much higher.
I commented on the post asking where the figure came from (in case it was from somewhere other than the graph and I was just misreading the post as attributing it to the graph) and he replied saying it was from the graph, but didn't reply to my follow-up comment querying the issues with it.
It could be that there's some formula or heuristic (with, necessarily, a lot of demographic assumptions) that I'm unaware of, that estimates a likely percentage of single-parent households from a percentage of single adults who have children, and that feeding 14% into that gives 33%, but he doesn't mention anything like that at all.
Re the election results. Journalists, particularly 25 hour news channels like CNN have an incentive to frame everything as close as possible because if they say the result the obvious people won't tune into their live coverage. This is fairly easy to do because both campaigns will also say it's close because that encourages their voters to turn out. Nate Silver has talked about this a lot in 538 podcast and his other writing.
Yup, that's a prediction market on whether other crypto investors will buy something. And the exchange is promoting this on Twitter with Pepe the Frog memes and suggesting you missed out on 7x profits.
I don't see how anybody could think that will provide useful information.
"This is near the top of the YIMBY movement’s policy wishlist - they seem to think most people would vote for denser zoning, though I have trouble understanding their optimism."
I don't think Street Votes is near the top of the wishlist. I think that's stuff like removing zoning. But it is near the top of what seems able to pass, especially in the UK. It's not that this will make most people vote for more density, but that it's a lot better than the status quo, where there are far more veto points. I think (80%) that in Israel street votes lead to significantly more housing, largely because it allows citizens to make a load of money having their whole street upgraded at once.
Family Feud also incentivizes contestants to pick popular answers.
It seems to me that this is just recreating the incentives that exist now for pundits--if you say what everyone else is saying, you're a sensible fellow who gets to keep being heard from; if you disagree with everyone else, you're some kinda nutcase who shouldn't be heard from anymore.
>I think the advantage of iterating it this way is that you can amplify small changes. Suppose that right now, the market thinks there’s a 50% chance that America will beat China in 2100. But I am a great economic analyst who knows things the market doesn’t, which allow me to determine that the real chance is 52%. Leaving money in a market for five years to make 4% return doesn’t sound great. So we could do something like make people bet on whether the 2025 market would be <40%, 40 - 45%, 45-50%, 50-55%, 55-60%, or >60%. This way if you’re even directionally right you can double your money in five years. I bet there are more clever mathematical ways to do this which would give you finer-grained resolution.
Look into https://www.investopedia.com/terms/l/leaps.asp
Essentially these are stock options with fairly high premiums dated anywhere from one to three years. These exactly solve your issue of amplifying small changes, as the price of an existing options contract automatically reflects changes in market consensus.
As an example, a Ford 01/24 20C will cost you $4.8 https://www.nasdaq.com/market-activity/stocks/f/option-chain in premium, while a 01/23 20C runs you $3.65. If you think Ford will go up more than decay by, say, 06/22 (fairly low over these time periods) you can buy one of these options and, if Ford goes up, sell for more than you bought it for whenever you want to, regardless of the actual expiry date.
The general answer to that issue is leverage, options is just one form of it.
What Scott is really talking about is best represented by an option fence https://www.investopedia.com/terms/f/fence-options.asp. Its and option strategy such that it pays out if an underlying security is within and range but pays nothing outside it.
With a liquid enough market in options with a wide range of strikes you can essentially define the market's view of the probability that an asset being in any range at a given point in time.
If you make the fence size small enough you approximate an Arrow-Debreu security which is a contract that only pays out in one state of the world.
https://en.wikipedia.org/wiki/State_prices
Wow that was a lot of typos. You could also get at something like what Scott is talking about (probably a little better) with a Butterfly options strategy as well. https://www.investopedia.com/terms/b/butterflyspread.asp
There's the possibility for a party-destroying primary in 2024 if Biden walks away. Harris is "in line" and is a treasured symbol to the activist identitarian wing that has taken over so much of the party; she's also one of the least popular politicians in the country. If she's nominated she'll likely be crushed by Ron Desantis or whoever, and it would essentially be a do-over of 2016 - nominating a deeply unpopular candidate who is constantly defended along identitarian lines. But if the opposition coalesces around Buttigieg then you'll get the Bernie Bro vs Hillary Stan war again, only much harsher. Not great!
How about "party-tanking"?
"An election or two" was your timeframe. And while it's hardly a settled issue within Marxism, Marxists have participated in bourgeois horserace politics for a long time.
There are certainly scenarios in which the Democratic Party in its current form ceases to exist. #1 is nuclear war, since political affiliation of the dead would massively slew Democratic and also they're typically in favour of cutting military spending which would discredit them. #2 is that the Democrats do something treasonous on the level of the Confederacy and get consigned to the political black hole in the aftermath (I know they weren't after the literal Confederacy, but AIUI it wasn't too far from happening).
I'm having a hard time seeing a line from the 2024 Democratic primary to #1, though. #2 seems slightly more plausible, but still far from clear.
Communists have run for office across Europe, with much success in the past. Particularly in France and Germany. The Die Linke party in Germany is a successor to the East German communists, and the Russian communists are the second party in Russia.
Marxists who don’t want to engage in bourgeois politics tend to be tiny sectarians, or entirely online.
Don’t they? I know Marxists in France are anti-Identitarian. I was responding to your jive about “bourgeois horse race politics” (elections)-communists do engage in that. And in the US too, with little success.
US online Marxists seem to lack any organisational power whatsoever.
> a treasured symbol to the activist identitarian wing
Is she? Basically all of the leftists I know, both online and in person, hate her because of her prosecutorial background and the fact that she’s fairly centrist. I’d predict that they would coalesce around Elizabeth Warren more than around her, though of course she’s in a great position to start with because she’s the VP.
Yeah in the primary she topped out at what, 5%? It doesn't seem like there's much of a constituency that's particularly attached to her. Being VP will help but I'm not sure it's enough to lead to the situation Freddie is describing.
I don't think many of the people who supported Biden in the last primary were particularly "attached" to him either. But he had plenty of establishment backing, which made him the default option. It feels like Harris is going to receive that establishment backing in 2024, and many of those same types of people will support her because of that, if not explicitly.
All the leftists I know, both online and in person, also hate her for her prosecutorial background and centrist economic tendencies, but I don't presume to think that the leftists I know are representative of democratic voters. The leftists I know mainly supported Bernie, with some going for Warren, or possibly Yang. Clearly they weren't representative last time. As of right now, if Harris wants the nomination, it seems like it's hers, unfortunately. Unless the field is small and progressives can unite behind another non-establishment figure.
My point wasn't that there's no way she could win, it's that if she runs and loses in a close race I wouldn't expect her supporters to be sore losers in the way Freddie was implying. That kind of support seems more likely to coalesce around whoever her opposition ends up being, although I have trouble seeing it for Buttigieg either- the "activist identitarian wing" didn't seem to like him much either. Ocasio-Cortez seems most plausible but who knows if she even runs.
Gotcha. I think people are using the terms "activist identitarian wing" and "leftist" differently here. I would not describe Bernie as an identitarian. To me, AOC fits the bill, but she's also got a good amount of support from the "economically progressive non-identitarian wing" - closer to what I consider "leftists". It seems to me like conservatives tend to conflate these two camps. Harris is obviously an activist identitarian, but I also think that what I consider to be the activist identitarian wing views Buttigieg favorably. As such, I think a head-to-head between Harris and Buttigieg would be very different from the Bernie vs Hillary war. Not much ideological difference there.
Hard to predict who will represent the "leftist" camp in the 2024 primary. AOC is the main person who comes to mind, but I'm pretty sure she'd get destroyed in a general election. Pramila Jayapal and Sherrod Brown might be better options.
"Identiterian", he said.
I read that as Neoliberals, rather than leftists.
The "more *clap* female *clap* drone operators!" crowd.
The identitarian wing is very, very distinct from the leftist wing.
The people who love her aren't leftists. They are identitarian liberals of the type who dominate New York and other large blue cities.
Are there any people who love her? The primary results suggest not in significant numbers.
I need to start checking Metaculus more. I was hoping to find something about Rittenhouse on Polymarket, but it didn't come up. Watching the trial has been rough. I'm not looking forward to the news following the results of the trial.
See also https://metaforecast.org/?query=Rittenhouse . It's a search tool for predictions.
Tell me about it.
Prosecution fucked it, defense fucked it, judge fucked it. Pure clown shoes in trial form.
The prosecution had their work cut out for them given that pretty much the entire event was captured on video and clearly supported the self-defence argument.
The prosecutor was reduced to claiming that the defendant's desire to protect himself was unreasonable which strikes me as a tough sell.
There were other things they could have done; more charges than murder or negligent homicide.
I'm not a lawyer and I've only done enough research to not accidentally commit a crime with my guns, but I wouldn't have done what he did.
He's a minor, he crossed state lines with a gun, the gun was being carried illegally (Fuck off with that hunting bullshit, the judge himself was fine with that charge until the cameras were rolling), Brandishing, it's every crime you can commit by having a gun except for stealing it in the first place or concealing it.
The thing I have trouble with is attempting to disprove the self defense claim, because I don't actually know anything about what qualifies you. My instinct (and the lectures I got about it) tell me this shouldn't be a covered offence.
If I drove across state lines to the Texas Republican Convention and stood their strapped and shittalking until someone did something and then shot them, that is NOT self defense.
AIUI, the weapon was legally carried due to the length of the barrel, which disposes of *that* charge.
I am not sure what evidence you have that Rittenhouse was 'shittalking', but even if he were I am not sure that justifies a physical assault on him.
What is the relevance of 'crossing a state line' in this context?
Well it can't be a *state* level charge anyway, it would have to be a Federal charge.
He carried the gun legally, that's clear. Crossing state lines means nothing, it was a 15 minute drive anyway. Brandishing maybe- not sure what you mean by that. Just carrying an assault weapon around strapped to your body is not brandishing, in many states that is perfectly legal and people do it. If he was pointing his gun at people that may be an issue and the prosecution tried to make that argument.
There is no evidence of shit-talking that I'm aware of, but I think that could be relevant depending on what exactly you said. Of course the guy he shot was screaming that he was going to kill Rittenhouse and eat his heart and stuff like that...
"He carried the gun legally, that's clear. "
That's exactly the opposite of clear, which is the point of the judge's actual ruling. Even as a professional with decades of experience in criminal law and many hours studying that particular statute, he professed uncertainty as to what was or was not legal.
*Because* it is unclear, Rittenhouse cannot be prosecuted for it. That's pretty solid in Anglo-American criminal law. But "you can't be prosecuted for X" and "X is legal" are not the same thing.
See this twitter thread on the paper: https://twitter.com/NunoSempere/status/1458030798255296513 for why the two experiments are not talking about the same thing.
Essentially I think that what's happening is that for the group of randos in the first experiment, "predict reality", and "predict what superforecasters will predict" don't lead to different predictions. So """"reciprocal scoring"""" performs well compared to scoring against reality for the group of randos. But note that """"reciprocal scoring"""" in the first experiment is not actually reciprocal; the superforecasters get scored against reality, not against the randos.
This is really different from the reciprocal scoring in the second experiment, which is actually reciprocal.
Any thoughts or updates on this question Scott?: Will this question be mentioned in an Astral Codex Ten Post in 2021?
https://www.metaculus.com/questions/6554/astral-codex-ten-mentions-this-question/
The lesson I'd want to teach here is "it is impossible to completely predict something that can read your prediction before it does the predicted thing".
(This is part of probably the most important computer-science proof ever - the proof that the halting problem is unsolvable. The proof amounts to: suppose there's an algorithm X that for any input algorithm Y returns, in finite time, "yes" if Y halts in finite time and "no" if it runs forever. Now I make an algorithm Y that goes "feed Y's own source code into X, and do the opposite of what it predicts". By assumption X finishes in finite time for any Y, so Y will always be able to finish in finite time if X predicts it runs forever, and certainly it can spin the wheels forever if X predicts it finishes in finite time. But this is a counterexample to the definition of X, so X doesn't exist.)
This is a good description of my objections to Newcomb's problem. The existence of the perfect predictor and its prediction power is supposedly a given, but I'm not convinced it's even mathematically coherent, so it may be intractable to actually reason mathematically about.
A version of Newcomb's paradox in which the boxes were transparent would definitely have this problem, but the normal version doesn't because you don't have access to the predictor's prediction before you act. You could, in theory, physically enforce this; you put the predictor on Mars and run it ~simultaneous with the actual choice (so that light-lag forces the events to be independent) and, instead of having physical objects in the boxes, just agree to the payoff matrix of Newcomb's paradox based on the prediction and on your action.
Yes, I had exactly this debate on reddit very recently, which I summed up as:
I think this gets to the ultimate point though, which is that Newcomb's paradox is not a single problem but a family of problems whose answers are very sensitive to the specific constraints. If you specify a perfect predictor, then the question is inconsistent. If you reduce the predictor's accuracy to make it consistent, you now open possibilities for two-box solutions. You simply cannot close those loop holes via contrivances on the inputs, because all that does is entail complementary contrivances in the solutions.
So like I said initially, on average you're better off picking one box, but there simply must exist circumstances in which two boxes is the correct answer too. That is, such solutions simply must exist by logical necessity.
> If you reduce the predictor's accuracy to make it consistent, you now open possibilities for two-box solutions.
How so? The payoff matrix doesn't meaningfully change with arbitrarily high accuracy even if it's short of perfection, so the two-boxer is still expected to walk away with significantly less money.
"Arbitrarily high accuracy" is just as impossible as perfection.
Regardless, what you're looking at here is akin to zero-determinant strategies for the Prisoner's Dillemma. If you have any knowledge of the predictor's history and/or its internal workings, you have some non-zero chance of winning the two box solution.
> "Arbitrarily high accuracy" is just as impossible as perfection.
The original formulation gives boxes with the values of $1,000,000 and $1,000. Exactly what do you think is the maximum accuracy prediction possible, and how much does it change the math?
There *are* strategies where the player introduces randomness into their own decision, and this does indeed lower the predictor's accuracy correspondingly. Do any of these improve the player's expected winnings?
>If you have any knowledge of the predictor's history and/or its internal workings, you have some non-zero chance of winning the two box solution.
"Winning" is not defined as being theoretically able to choose to two-box and walk away with $1,001,000. If that non-zero chance is sufficiently close to zero, two-boxing is still a bad idea just as surely as betting on 00 at a roulette table.
Newcomb's paradox has no issue with the halting problem so long as you can't see Omega's prediction. Its prediction doesn't impact your decision at all, so it can be correct. If you can see what prediction Omega makes, than Omega is solving the halting problem unless you happen to make the same decision regardless. Maybe Omega makes predictions for what everyone passing by would do in each scenario, and only opens the door for those who's actions wouldn't be inconsistent?
> Newcomb's paradox has no issue with the halting problem so long as you can't see Omega's prediction
This isn't correct. The Halting problem isn't undecidable just because the program can change its behaviour based on the prediction, that's just the simplest degenerate case. Goedel proved that there are in fact an infinite number of undecidable propositions. Undecidability is fundamental and inescapable.
Since a human can simulate a Turing machine, then Omega must have its own set of undecidable propositions for which predictions simply must be considerably less than certain.
"Since a human can simulate a Turing machine" [citation needed]
But it's easy to get around that. Just add a provision that says you need to choose within 10 seconds, or else you automatically 1 box it. Though realistically, unless you never choose a box due to an unterminated loop or something, the halting problem shouldn't enter into it anyway.
I do wonder about the incentives going into those expert predictions in Short #2. Do any of them get any benefit from being right?
And the distinction between "good forecaster" vs "academic" vs "pundit" seems important too - so-called "experts" are quite often political animals rather than truth-seekers.
That is to say, in solid agreement with your comment, that in many cases the "experts" aren't even trying to make accurate predictions
> (this is just the endogeneity problem, but for the future instead of the past!)
Put differently: It would seem that using conditional prediction markets to make decisions might turn you into an evidential decision theorist...
> "Too close to call", reciprocal scoring, Theranos, Ritterhouse, long-termism
Rittenhouse doesn't seem to be mentioned in the body of the post at all?
> Rittenhouse doesn't seem to be mentioned in the body of the post at all?
The box above "I was wondering when this was going to show up"
Oops, thanks, seems I missed that.
> turn you into an evidential decision theorist
Wait, I was just wondering if this considering this problem was how Robin Hanson came to think of futarchy. Doesn't the commitment to picking the choice predicted to be better turn it into causal decision theory?
I don't think so. In the example, policymakers are trading off the benefits of a mask mandate against other values. The benefits are larger when the pandemic is very bad, so even if you commit to picking the choice predicted to be better you are still more likely to implement mask mandates if the pandemic is bad. So the same problem occurs.
I'm talking about a changed legal environment where what we now call the authorities have no choice of policy. Their job is to implement the decision which the prediction market made. And this is common knowledge of the participants in the market. https://en.wikipedia.org/wiki/Futarchy
Yeah that's what I'm talking about too, that's what I meant when I said "if you commit to picking the choice predicted to be better". Maybe you thought I was talking about a scenario where politicians say that they are going to follow the prediction markets, but still have the ability to do something different?
The opposite. I model futarchy as: the market estimates (Y1 deaths | do(mask mandate)) and (Y2 deaths | do(no mask mandate)); then the state mandates masks iff expectation(Y1) < expectation(Y2). In this model the decision node (mask mandate? or not?) is screened off from any nodes about the *reasons* for the decision. The betters will have a variety of models of the situation but they'd all share this property (assuming they trust the institutions to work as claimed).
Actually, the state will mandate masks if the difference in deaths is more than the costs of the mask mandate policy in terms of other variables that the state has pre-defined. If the pandemic is bad then the deaths saved by the mandate are more likely to overcome this threshold, even if the mandate saves lives regardless of how bad the pandemic is. So the state is more likely to implement a mask mandate if the pandemic is bad.
Re: “Too Close To Call”
The networks basically have three classifications (in a 2-person race): Called for A, Called for B, Too Close to Call. They don't want to deal with things like "favored", and given what we saw in the 2020 presidential election, I'm inclined to agree with them. The NBC and ABC decision desks won't call a race unless they're at least 99.5 confident in the winner. PredictIt and Polymarket weren't at that point yet, although prediction markets can be a little wonky around the extremes.
NBC: https://www.nbcnews.com/politics/2020-election/nbc-news-decision-desk-how-we-call-races-election-night-n1245481
ABC: https://fivethirtyeight.com/videos/the-decision-desk-wont-project-the-winner-of-a-state-until-its-99-5-sure/
Yeah Scott seems to be missing the fact that in this context "call" is a technical term meaning announcing that a candidate has won, not just saying "this candidate seems to be winning." 98% just *is* actually too close to call by this standard.
Yes, but giving people lg(101) bits of information is way better than giving them lg(3) bits of information.
But you're not giving people log(101) bits of information. Information is not conveyed magically from the mind of the reporter to the mind of the reader.
Treat the news as a noisy channel, with reduced information capacity. Headlines magnify that problem since informational (non-clickbait) headlines try to convey information without the benefit of the full article.
I mean reporting a percentage 0-100%, like the betting markets do, is lg(101) bits of information. News reports compress that range down to just three options: Yes, no, and "too close to call". I'd be happy if the news just quoted the prediction markets.
This is why the needle is so great!
In particular, extrapolate it to a 50-state presidential election. If you call each state when it's 98% sure, your chance of getting them all right is .98^50 == .36, so 64% chance at least one is wrong. Let's be more realistic: plenty of states never get as low as 98%. Keep it to the 10 or so "battleground states": now 98% right gets you "only" a 18% chance of getting one wrong. Hope that one wasn't Florida 2000!
In addition, if you read election day articles, or watch election day TV news coverage, the analysts often say things like "Candidate A is favored", or "Candidate B has an increasingly narrow path to victory. These aren't exactly probabilities but tell the same story as them.
This is the correct answer. See also the Fox (and eventually AP) call of Arizona in 2020, which was predictably correct at the time it was made but nowhere near the 1/200 threshold it should have been and therefore a bad call.
“Too close to call” and “too early to call” are not exactly synonyms though.
Scary detail in that "Expert forecasts vs. reality" tweet that's easy to miss (and complicates the story of systematic mis-calibration a bit): up through 2008 or so the life expectancy projections consistently erred on the low side. Since then they've consistently erred on the high side.
Related: When one sees these instances of experts repeatedly predicting a trend change and repeatedly being wrong, one is inclined to lambast them saying "They should have seen all these previous predictions of a similar trend change, and how they had always been wrong, so maybe they should have learned the lesson and they should have just predicted that the trend would continue." But in what time frame should they look at the trend? After 2008, should they predict that the long-term trend (one of relatively fast increasing life expectancy) will continue, or the short-term trend (of much slower increase) will continue? Or maybe they should predict that the short-term will continue *on the short term* (but neither prediction is obviously right on the long term)?
In fairness to traditional media, part of the issue is that their standards of proof are super high. They make hundreds of predictions per election, and they get one wrong maybe once every decade or so. They probably do know that the chance of a particular candidate winning is 98% (or whatever), but by their standard that just isn’t good enough, because that would lead to many more missed calls.
That's the idealistic take. The cynical take is that they just want you to keep watching, and as soon as they call it one way or the other, half the audience is going to change channels.
Honestly it’s more realistic to say that they just are really conservative on calling it. If they get even one major call wrong they’ll get lambasted for weeks, and no one is materially better off for knowing the result a couple hours sooner.
I'm guessing this is what applies to actual "calls", while the keep reading incentive drives horse race coverage in the preceding weeks.
I have seen the third graphic of XiXiDu from people who claimed that the graphic is nonsense and that XiXiDu got it completely wrong. The many horizontal lines are not predictions. Rather they are summed up plans of governments of various countries on how much solar capacity the governments plan to fund in the next years.
I didn't verify the claim, so if someone can say something about it (or about the other three graphs), that would be nice!
What does the prediction markets say about Trump running for - and winning - the Republican nomination for US president in 2024?
I ask, since back in 2020 I betted a bottle of good italian wine plus a dinner at a nice italian restaurant that he will run & get the nomination in 2024. (Hedging, really)
More generally, I am worried that if none of the sane US political people pick up obviously sensible policy proposals, like immigration control of irregulars & compensating those who lose out of economic globalisation, then a window opens for the insane to take power. Like in Russia 1917 and Germany 1934. Not as dramatic as there, but still.
May sanity prevail among the not-insane in the US. It is important also for Europeans, and everyone else for that matter.
Metaculus has it at 55% https://www.metaculus.com/questions/5049/if-trump-is-not-sworn-in-as-president-in-2021-will-he-become-the-republican-nominee-for-president-in-2024/
PredictIt shows a 40% probability that Trump will win the Republican nomination. But this far out, I wouldn't give it a lot of credence. Primary elections are hard to forecast because the candidates are so similar ideologically.
Thanks Chipsie and Anna.
My own subjective estimates, for what they are worth, have been 50/50 that he will run, 70/30 that he will win the Republican nomination, and 40/60 that he will win the election. Which gives a probability of .5x.7=.35 that he will win the nomination (enough to bet a bottle of good wine), but only .5x.7x.4=.14 that he will become the next US president. (For Trump himself it would boil down to .7x.4= .28 though.)
All subjective estimates are basically talking-through-one's-hat (as the English say), but on the other hand you often need distance to see things clearly. I never understood my own country better than after having lived a year in the US (the fog lifted after approx. six months), and maybe it works the other way as well.
Seriously, all the havering from the Democratic side about Trump makes me think they are building him up as a credible candidate in 2024.
While it's to their short-term advantage to keep scaremongering about the Trumpian menace and likely coups, in the long-term they are contradicting themselves about him being powerless and ineffectual and reduced to angry outbursts on social media from his exile in Mar-a-Lago when they and their supporters write lurid pieces about plotting coups:
https://www.theguardian.com/us-news/2021/nov/14/trump-president-2024-election-coup-republicans
Good point.
We are in the middle of the defining-the-situation political game (which is at least 80 percent of what media-led political discussions are about, everywhere), and it is always dangerous to run two conflicting narratives/definitions of the situation at the same time (Trump is cunning and clever/Trump is a nutcase).
I have a hard time making up my mind myself. Sometimes I believe he is both, which is the most dangerous kind of politician. I actually hope he is ”only” cunning and clever, since I would much prefer a clever half-crook to occupy the White House (like Nixon), than somebody who - insanely - really believe that the election was stolen from him (plus the old Obama birth certificate nonsense).
With the benefit of hindsight, claiming “the election was stolen from me” is rather clever, if he plans to run again in 2024. The US dislikes a loser. Running like “ last time’s loser tries a second time” is unlikely to cut it. Running like “last time’s winner whose victory was stolen from him tries a second time” is a better narrative, from Trump’s point of view. Not much better perhaps, since he must convince the median voter (who hopefully prefer a sane ruler) that he is not insane enough to actually believe so himself, and only pretends to believe this, because he is very clever and knows it positions him better for a re-run…the message is rather convoluted.
However, if the Democrats mess things up sufficiently, with lack of immigration control, no protection of the US working class against competition from abroad, the likely very costly “green shift”, and fumbled foreign policy, enough voters may hold their nose and take a chance with him again in 2024.
A better strategy for the Democrats would be to instead "steal" many of Trump's policies in the years ahead (packaged a bit differently, of course). If they fail to do this, I’d adjust my prior of him winning in 2024 to 50/50 (if he runs)😊.
But isn't Trump actually great for them, compared with somebody who is both opposed to them and might have the ability to do something about it? Trump had the majority in both Houses for two years and the only thing he managed to do was slashing some taxes, the apocalyptic-level menace that he is.
If Biden does no better on COVID (and despite a vaccine it looks pretty similar to me - the restrictions are still in place; more people have died of COVID with Biden in office than Trump), and there are other things that bother people (inflation, Afghanistan, immigration, whatever) that is still an issue approaching 2024, then I think a lot of people will remember the Trump years differently than they would have said at the time. Biden ran as a centrist uniter, but hasn't really governed as such. Trump lost in large part because of COVID, otherwise the very close election could have easily, I would say would have, gone Trump's way.
Trump's years, outside of the Russia allegations and the tweets/self-aggrandizing lies, were actually pretty calm and prosperous. With the Russia allegations looking worse for the media sharing them, Trump's personality and weird lies are the last major issue. People tend to forget over time, or remember more fondly, concerns of the past. With your point that Trump didn't change a whole lot, it may sound to many like a fine alternative in practice, even if Trump himself looks like/is a mess.
Trump is less likely to win than a moderate Republican, which means he's a better opponent to have in that sense. But the downsides of him winning are larger. In particular if you are concerned about the erosion of democracy, a small chance of never being able to win an election again is worse than a higher chance of losing one now
Sure, if those are the two options. What the establishment is actually afraid of, it seems to me, is that a much more competent Trump might appear, a populist authoritarian who has a realistic chance to bring about that erosion. If there's an actual demand for this sort of personality on the right, a prominent buffoon already in place seems to be a lesser risk than a vacancy which might get filled by an unpredictable dark horse.
The IEA estimates have a huge bias against r
Solar. I won’t speculate on why.
(That said I read, independently of this, their reports on wind which was quite gung ho).
To be honest, I don't see how reciprocal forecasting is supposed to work in game-theoretic sense. Normal prediction market has nice and obvious Nash equilibrium: just predict the probability as best you can. Not so in reciprocal forecasting. From what I can tell, a strategy closest to optimum is to shout at loud as you can "Let's all vote for X" and then vote for X. If it is even possible to do this without communicating with others by forecasting the most trivial outcome, like answer 50% to any question.
Iterative prediction is far more plausible. You don't even need to do the iteration if you ensure two conditions: 1. The stakes are freely tradable at any time. 2. While the prediction stays the same, the value of the stake is growing at least as fast as some default assets like S&P 500. If the second condition is not fulfilled, then unless the market is extremely wrong, long-term it is more profitable to invest in other asserts. This can be achieved by, you guessed it, having the market operator invest the stakes in that very investment product. This could even happen automatically by making the stakes in some cryptocurrency like ETH, if you believe that ETH will continue growing at least as fast as S&P 500.
It's just a Keynesian beauty contest. You have a strong incentive to coordinate with everyone else, not to be right.
2) is possible to achieve by availability of leverage. For example, for any sufficiently liquid base contract, one can set up a futures market with risk based margining. Then each long or short participant can invest most of their cash in any asset mix they like at each point in time. This may be easier to implement than to mitigate credit risk on a market investing market operator or agreeing on a single default asset.
Biden himself has said some things about being a one term president. Mixed with his age and his low approval ratings, it's not surprising that the markets are looking for alternatives.
It's a sad state of affairs that the candidates available are as poor as they are. I don't think any of the listed candidates can beat Trump, let alone a more solid Republican that doesn't make half the country reflexively convulse at hearing their name. Those that excite the base scare the middle, and vice versa. The fact that Hillary Clinton appears in the group is certainly not a great sign for Democrats. It's early yet, though, and maybe some new candidates or a different consensus candidate can still emerge. Or again, maybe Biden turns things around and doesn't get any health issues and decides to run again. Obviously I'm not expecting that to all happen. Even if he does turn around his approval ratings, that might be a great time to retire with a good track record and leave a legacy, rather than chancing that he'll make it to 85 in one of the most grueling jobs in the world.
Biden would likely lose right now to most Republican candidates, though he has a few years to potentially turn around his unpopularity. Harris is riding her own unpopularity as well as Biden's. Interestingly enough is the recent soft takedown articles from left-leaning mainstream news sources. They don't seem to like her either, and may be paving the way to have another candidate in 2024. Buttigieg is not popular with the farther left, and most of the others are multi-time losers who are even older than they were the last time they ran. AOC is too inexperienced to run a nationwide campaign, and too limited in her popularity to unite the Democrats. She's a darling of the far left and young people, but unless she comes sharply to the center, she's going to scare away a lot of the middle and independents.
I read that list as Democrats being in real trouble in 2024. Trump could be divisive enough to give some of them a chance, maybe, but a better candidate seems like a pretty obvious win for Republicans at the moment.
I think a lot of that list represents unknown unknowns - a similar list in 2013 for the 2016 election wouldn’t have listed Donald Trump at all!
People think that Biden won’t run and Kamala won’t win, so they want to sell Biden and Kamala and without any other reasonable options yet they just list famous Democrats. I don’t think there’s a chance that AOC will make a serious bid, Bernie won’t run because of his age, and Clinton definitely won’t run either.
Buttigieg and Warren are both realistic options (and correspondingly get some predictions on the list) but it’s so far away from the election that it’s still pretty meaningless.
Buttigieg maybe, in time - he's currently Secretary for Transport so he is getting that experience of being in government under his belt, but he's still young and inexperienced (for a politician) and he's not wired into a political machine like Obama and Chicago.
I think Warren has shot her bolt - she had a chance last time, but since Biden was always going to be the pick that went nowhere, and she kind of has "a great future behind her" - she was touted as going to do great things, but she's been around now for a while and that kind of candidacy never really did happen. I think, to be frank, that by 2024 she will be too stale.
I have no idea who a potential Democratic candidate that is not Harris or Biden might be, though.
"a similar list in 2013 for the 2016 election wouldn’t have listed Donald Trump at all!"
Yes, the most amusing thing that came out of the 2016 result were the victory videos showing all the pundits confidently predicting "Clinton will win this state, she will take that state, the polls say such-and-such" and on the night, those states went for Trump.
https://www.youtube.com/watch?v=fT0AjmeJ_sg
I think everybody enjoys seeing pundits ending up with egg on their faces. The sniffy certitude on display that of course Hillary had it in the bag makes it all the sweeter.
As well as everyone from Lin-Manuel Miranda up to Obama saying "Never gonna happen" and then it did.
https://www.youtube.com/watch?v=XqEddipbpkw
I definitely agree that it's too early to rule out other, new candidates. The problem is Biden sucking up the official oxygen by being a first term president. To actively put themselves out there when he hasn't said whether he plans to run again risks going against the party. If he does decide to run, and presuming he hasn't fixed his approval ratings sufficiently, then potential candidates are going to have to really think about whether to run anyway, instead of sitting things out. Tough choice all around, unless he backs off and makes way.
I definitely agree that if he chooses to run for re-election then he’ll almost definitely win the primary, probably by default.
If it does turn out to be Biden versus Trump in 2024, I have no idea what will happen.
You would expect that selecting Kamala Harris as Vice President was giving her the chance to build her candidacy for 2024, but there are some odd little stories here and there which whisper at the rift within the lute and dissatisfaction (allegedly) with her performance to date:
https://edition.cnn.com/2021/11/14/politics/kamala-harris-frustrating-start-vice-president/index.html
Presumably that is why some people are betting a different candidate than Biden (on grounds of age) or Harris could be selected.
It's funny you say Hillary Clinton being there is a bad sign for Democrats -- I was thinking it's a bad sign for the reputation of prediction markets. There's something about Hillary that drives ~10% of bettors insane, which the only reason her name is even at 2c up there.
Assuming I were 100% confident that it goes to zero, I'd have to spend $980 right now to make $1000 when the contract resolves in 2.5 years, which turns out to be less than a 1% return.
Scott was talking about trying to bootstrap returns from far future events, but there's only so fast you can bid down something with this small of a return.
It's not about making money, it's about prediction markets as a means of finding truth, which is what Scott is advocating for in this series. Hillary has nowhere close to a 2% chance of getting the 2024 nomination. She had nowhere close to a 10% chance of being arrested and jailed. She had nowhere close to a 10% chance of getting the 2020 nomination (nor did Yang, for that matter). But those were the chances that the PredictIt markets settled on. You can't expect people to trust odds set by prediction markets when those markets are frequently distorted by cultists and conspiracy theorists.
I also agree with that, but the prediction markets would have found some other names to say instead if there were realistic candidates. That speaks to the idea that they don't think Biden will be the candidate, but really can't think of someone who makes sense that is actually plausible.
A lot depends on who emerges into leadership, on both sides, after the Democrats are slaughtered in November '22.
I'm curious to see if the Reddit prediction feature works. Getting prediction tokens seems like it would be good for clout in a subreddit, but the strength of that incentive probably varies a lot across Reddit. Too bad PredictIt and PolyMarket don't have subreddits. If Reddit can beat a PredictIt market on a question that matters, that would be really surprising, because seeking clout on a subreddit would be a better incentive than money to reveal hidden information and make accurate predictions
>I’ve heard a lot of stuff about the prosecutor really bungling this one, but mostly from conservatives who I would have expected to hate the prosecutor anyway, so it’s good to get objective confirmation that yeah, this isn’t going anywhere.
I'm not saying the prosecutor is doing a great job, but the main disconnect here is between "he did something bad" vs "he did something illegal." The law sets out specific criteria about what counts as self defence, and I think Rittenhouse both was irresponsible and definitely hits the bar for self-defence here.
Think of it as similar to QI things, where people are shocked cops aren't convicted for doing bad things. We wrote laws/made SCOTUS decisions saying you can do *extremely* bad things without it being a crime. I don't think the prosecutor screwed up, I think he's unconvictable on the merits of the law.
I agree... but then why proceed with the case? If there's no reasonable prospect of conviction, he bungled it by trying the case
I think because it was so controversial and trying the case means people can see the adversarial process play out and hypothetically have peace with the outcome, vs it being the prosecutor’s private call.
And this is very, very important - that justice is seen to be done is arguably more important to public order than actually having just results.
One possibility is that there's room between "no reasonable chance of conviction" and "conviction at trial is likely". If the prosecutor judged something like a 20% chance of conviction at trial, it might make sense to proceed anyway. If the prosecutor believed Rittenhouse deserved to be punished (either because he's likely in the prosecutor's judgement to be factually and legally guilty even if not necessarily to a reasonable doubt threshold, or because the prosecutor is assessing dessert based on abstract notions of justice or cynical judgement of the needs of public order rather than a faithful adherence to the letter of the law), a 20% chance of conviction is better than a 0% chance if you don't bring charges.
Moreover, the threat of going to trial when there's some chance of conviction but not a great chance is how prosecutors get leverage for the plea bargaining process by which they secure most of their convictions, giving a cynical reason to prosecute a trial you're likely but not certain to lose. It's like being the Dread Pirate Roberts: once word gets out you've gone soft, it's nothing but work work work all the time.
Both of the above are compounded by the possibility that the evidence looked better to the prosecutor before trial. Key testimony from prosecution witnesses seems to have come out more favorable to the defense at trial than the prosecution was expecting, and also the judge made some rulings on evidence admissibility that were likely more favorable to the defense than the prosecution was expecting.
Even if the prosecutor did believe he had no realistic chance at conviction, there are plausible cynical reasons for going to trial anyway. Both for public order considerations and for his own reputation and career.
Going to trial could be an exercise in failure theater. If the prosecutor dropped charges before trial, it's look to members of the public who believe Rittenhouse to be guilty like the prosecutor (and the system he's part of) was complicit in letting a murderer go free. This make the prosecutor himself a villain, further undermines shaky confidence in the system, and risks triggering more violent protests in response to the perceived injustice. Taking the charges to trial gives a chance of shifting perceived blame from the individual prosecutor and the legal system as a whole to the judge and the jury.
There's also a potential angle of "the process is the punishment". Facing murder charges and the possibility of life in prison for a year or so isn't nothing, even if it's enormously preferable to actually being convicted. If the prosecutor believed Rittenhouse deserved punishment, or the public order required Rittenhouse's behavior to be disincentivized, then putting him through the rigours of trial is one way to do that.
Political aspirations of the prosecutor.
Apparently, the curfew charge was dismissed because the prosecutor *literally forgot to present any evidence for that charge*.
Interesting. My impression was that they knew they couldn't win on it (due to a lack of any effort at enforcement), and they brought it just to imply that he was there illegally and add to the perception of his recklessness / lawlessness; and actually arguing the charge would have undermined that perception.
Same with the gun charge, finally thrown out today. The statute in question doesn't apply to the gun he was carrying (not a "short barreled rifle"), nor to him as a 17 year old, I think. (I believe it has an exemption for 16 and 17 year olds.) But the prosecutors brought both these inapplicable charges to better hand-wave at "illegal gun, illegal presence, reckless conduct". Interested if you have a source contradicting that.
The gun statue is confusingly worded, and both sides think that it obviously applies or doesn't apply to Kyle. From what I've heard, the judge threw it out precisely because it was so confusing.
As for the curfew charge, the impression I got is that they'd have had a slam dunk if they bothered, but I'm not a lawyer, so I have no idea how those things actually work.
Isn't there a problem with the curfew charge, that there were literally hundreds of people actively breaking the curfew, including the people who Kyle shot/was being chased by?
Not very sympathetic to go after Kyle only on a curfew charge. I also find myself wondering if the lack of prosecution of the guy who got shot in the arm for having an illegal weapon there may influence the jury?
I've heard people say it's confusing but the gun statute seems pretty clear.
https://docs.legis.wisconsin.gov/statutes/statutes/948/60/2/a
If the gun is a rifle or shotgun the charge doesn't apply unless it's short barreled or the individual is younger than 16 and in violation of 29.593. And since he's over 16 and the gun isn't short barreled it seems like he can't be in violation of the law.
The language is "not in compliance with 29.593", not "*in violation* of 29.593". Those aren't necessarily synonymous, because it isn't clear whether one can be "in compliance" with a law that isn't relevant to the issue at hand. Is a pilot flying an airplane over Milwaukee "in compliance" with Green Bay municipal traffic regulations? If you think the answer is obvious, then either you know more about Wisconsin law than e.g. a Wisconsin judge, or you don't know as much as he does.
The legal arguments I've seen that he was in violation of the law weren't arguing the case you made here but a more complex case around legislative intent. Basically that the law as stated says he's not guilty but that was a mistake of the legislature.
Having watched an unhealthy amount of the trial footage, I agree with your overall assessment, but I think it is still soft on the prosecutors.
In particular, ADA Binger's antics, including trying to bring in excluded evidence and questioning Rittenhouse about remaining silent pre-trial (a big 5th amendment no-no), led to some dramatic moments last Thursday. It got to the point where Binger gave his rationale for his actions and said they were in good faith, and the judge looked him dead on and said (inexact order of phrases), "I don't believe you... you're an experienced prosecutor. I don't believe that was in good faith. I'll take the motion [mistrial with prejudice] under advisement. And I don't believe you." I think it's clear that he's hurt his own case with his shenanigans, quite possibly to the point of invalidating a guilty verdict if one is delivered.
That said, I've seen people claiming that the prosecutors should receive professional discipline, or be tried for misconduct, and stuff like that, which is ludicrous. As the saying goes, they can't pound the facts, they can't pound the law, so they're pounding the table. There's nothing punishable or even remarkable about it.
I haven't watched any of the footage, but my first impression of seeing a photo of Binger was "Why does he have a Tintin haircut?"
ADA Binger: https://nypost.com/wp-content/uploads/sites/2/2021/11/judge-rips-rittenhouse-prosecutor-01.jpg?quality=90&strip=all
Belgian reporter and adventurer Tintin: https://static.wikia.nocookie.net/tintin/images/d/d9/Tintin_TV_Show.png/revision/latest?cb=20190623011522
And apparently he wears Star Wars lapel pins?
There seems to be a combination of "DA realised this thing was toxic no matter what way it goes, so he handed it off to Binger" and "Binger is ambitious, ran for DA in another county, and wanted this as big case to make his name" at play.
> I don't think the prosecutor screwed up, I think he's unconvictable on the merits of the law.
Then the prosecutor did screw up for laying those charges the begin with. They try really hard to only go to trial when there's a chance in hell of winning.
There was a chance in hell of winning. He was unconvictable on the merits of the law. But he was convictable by jury intimidation.
Yeah, there have been some “bungles”, but when the most significant “error” is allowing your alleged victim to described what happened to them, you probably didn’t have a great case to begin with.
It did get bad enough that some people (including the judge!) implied the prosecutor may have been angling for a mistrial. But that was because the prosecutor was trying to slip in some disallowed lines of argument (again, less an error than an admission of a weak case).
Another in the "experts get things repeatedly wrong" category: all the big investment banks have teams of analysts who try to predict how much profit public companies will make in the future. It is well known that these forecasts are consistently wrong -- specifically they are upward biased. Interestingly this bias persists despite there being a significant financial interest in getting the numbers right, since funds often rely on these figures when making stock picks.
One theory for why this is so is that the analysts intentionally publish bullish forecasts to flatter company management, which helps them e.g. get meetings with the executive team.
Well. That makes sense.
I doubt anybody puts much faith in them, given that there are already prediction markets for how well a publically traded company will do (aka markets)
Maybe they should hire people with consistent moderate depression to temper the optimism.
>generally avaliable
>Right now it’s only in select areas
Isn't there a bit of a contradiction there?
>and decided the best paper out there was Brauner et al
I'd question whether these kinds of questions can be measured by science (or superforecasters) at all, but I'm a bit confused that "closely correlated estimates" is supposed to be impressive. The point of the estimates isn't to be closely correlated, but to take approximately the same values, no?
The "select areas" are fairly widespread, but focused on regions that don't already have great internet, which explains why California isn't a priority. Even the best Satellite internet will still have fairly high latency, which does limit its usefulness if you have access to fibreoptic lines.
Addendum, I actually went and looked it up, and the current satellites cover the USA, Europe, Australia, New Zealand, and Chile, as best I can tell. Not quite sure what Chile is doing in that list tbh, the rest are all wealthy countries. Also not sure why bay area isn't available yet if there are satellites covering it...
"Their solution is to make teams very big and full of smart people, so that it’s unlikely other teams will miss something."
Because I am an incurable smartarse, I can't help feeling about prediction markets as per "The Napoleon of Notting Hill":
“The human race, to which so many of my readers belong, has been playing at children's games from the beginning, and will probably do it till the end, which is a nuisance for the few people who grow up. And one of the games to which it is most attached is called "Keep to-morrow dark," and which is also named (by the rustics in Shropshire, I have no doubt) "Cheat the Prophet." The players listen very carefully and respectfully to all that the clever men have to say about what is to happen in the next generation. The players then wait until all the clever men are dead, and bury them nicely. They then go and do something else. That is all. For a race of simple tastes, however, it is great fun.”
RE: the media all claiming the Virginia race was too close to call - well, the final result was Youngkin with 50.6% of the vote and McAuliffe with 48.6%, so it was pretty close. But my cynical interpretation was "These are liberal to left media news organisations, naturally they will be biased towards 'No, it cannot be that the uncouth Republican will triumph in Blue Virginia, so we will not give the oxygen of publicity to these threats to American democracy".
https://www.270towin.com/states/Virginia
https://www.washingtonpost.com/blogs/erik-wemple/wp/2017/01/27/dear-mainstream-media-why-so-liberal/
https://www.theatlantic.com/ideas/archive/2021/04/gop-grave-threat-american-democracy/618693/
They were pretty sure Biden won the 2020 election, but were hesitant to say so there, as well.
As has been pointed out, it's more that all their incentives point towards not calling any race until they have no choice. A) viewers want to watch a race, not a victory lap. B) being right gets you nothing, but calling an election wrong gets you excoriated
I'm going to push back on Erik Hoel's prediction "#17. People and culture will become boring." TLDR: Because it will be possible for a single individual to create entire movies or series in much the same way a single person can create a book right now, so we'll have an explosion of new ideas in media.
He argues:
>The panopticon of social media and state control will lead to cultural stagnation. We already see early hints of this. Consider the remakes of older movies: 2050 will be a stew of remakes of remakes, and familiar and boring intellectual property (like Star Wars) will be king. Creativity vivacity will suffer, especially in the arts and humanities. The coming half-century will be a great one for innovations in finance, engineering, space travel, and artificial intelligence. It will be a terrible one for the arts and basic scientific advancements (like a new physics), for such advancements require iconoclastic and creative lone individuals. This prediction is already augured by judging the 2000-2020 creative period overall in areas like art, music, literature, film, and scientific discoveries, and finding it severely lacking compared to, say, 1950-1970.
I think the reason we've seen less creativity is for economic reasons, not cultural ones. It's just so much money and work to make (and market) a good movie that studios play it safe and mostly stick to works they know will be an economic success.
But by 2050, AI will have made it so a single hard-working individual will be able to create an entire movie by themselves. The computer will render the scenes described and the 'author' will be able to direct changes until their creative vision is fulfilled.
This will make movies like youtube circa 2008 when there's all kinds of new opportunities for creative people and tiny little indie director/authors will go viral. Eventually the movie media will look more like our print media. There will be well established author/directors making a new movie every couple of years in far more genres than exist in current cinema. And it will be much easier to be profitable with creative niche projects so more people will be willing to try them or create them for art's sake or as a hobby. If you only need 20,000 people to purchase tickets to a movie to make it profitable, it opens up the room considerably for more creative media.
At the same time, VR will open up a whole new space for new experiences. It's not a coincidence that early cinema was an age of creativity as it takes a lot of experimentation to figure out what works. VR is going to go through the same process as it slowly becomes a major source of our media.
And so eventually, we'll consider the years from around 2040-2060 to be a golden age of creativity.
I am very dubious about the creativity potential of the average guy. What exactly came out of that era in 2008 YouTube anyway?
You don't need the average guy to be creative. You only need the top people. If it's possible for the top 1% of creative people to make a living doing solo movie projects and only 1% of those people decide to make movies for a career that's ~50,000 author/directors. And potentially 10s of thousands of 'indie' movies coming out per year. And you'd only need the average American consuming a few of those movies (as paying customers) to make it profitable, which seems plausible.
Early YouTube just had a huge variety of creators all testing out the platform. So there was lots of variation and creativity.
I’m very dubious about the top 1% of people creative enough to produce worthwhile art - that’s close to 3 million people in the US. The number of great directors alive is closer to a dozen.
And are we in a golden age of art right now because of new technology? What art has been solely produced by YouTube?
And since mass literacy everybody has has the ability to become writers, the result hasn’t been an explosion of great writers relative to population, but the reverse.
Freddie Wong and Corridor Digital would be my go-to examples of indie filmmakers on Youtube. They've made some very impressive shorts with a very small crew and budget.
(Corridor has a bunch of behind the scenes videos, and probably half of them mention how a traditional production would have done [expensive thing] to get a certain look, but they've substituted some VFX magic instead.)
"VFX magic" often isn't, unless it's done sparingly and very, very well.
And while someone might be willing to overlook flaws in an amateur, small-scale project, when that is now allegedly a professional production vying for paid market, they will expect more and be less tolerant of "you could see this was greenscreen!"
I generally can't notice on a casual view, I would have to pause the video and spend a couple minutes thinking "should the shadows *really* be falling in that direction?" to figure out what's been composited. And they have gotten into the paid market (not just ad revenue, like, actually paid to produce a show), which is a more objective measure of competitiveness.
But my point was more about the budget - the OP's prediction was that improvements in CGI software and AI would allow a single author to create a whole video by their lonesome, and Corridor's shorts are what I'd call the first steps towards that vision. It's nine guys instead of one, and the cameras and gear aren't exactly cheap, but it's getting better all the time.
Yeah, but if I go onto Apple Store or Netflix or wherever, there are a *ton* of movies I heard nothing about, had no idea were even produced, and I have no interest in watching. You're forecasting an explosion in even more amateur movies competing for people's attention.
How many blogs are there? How many Twitter accounts? How many Instagram influencers competing for money and attention? A few will be successful but the majority will be scrabbling for pennies (relatively speaking). It may or may not be a burst of creativity, but getting the attention of even 20,000 people for your one project is going to be exponentially more difficult due to those tens of thousands of indie movies all wanting their eyeballs. You're more likely to fragment the market, with maybe 1,000 people paying to stream "My Brilliant Take On Noir Detectives In Space" over the Internet, rather than 20,000 willing to pay for movie tickets in a traditional theatre and going out to watch that movie. And the prices will have to drop correspondingly, who wants to pay you $10 when they can watch all that free content others are churning out?
What you describe has already happened for writing. Anyone can write a novel, self-publish at no cost, and have it available to everyone on Amazon. Has the quality of fiction gone sharply up as a result?
It's true that writing a novel was always something that an individual could do but he also had to get a publisher to publish it, which was a pretty severe filter. So not exactly equivalent to your case, but in the same direction.
Good point, but it's not the ability to publish that I'm thinking of so much as the ability to create in the first place. One person has always been able to create a book by themselves and I think in ~20 years one person will also be able to create a film by themselves. They'll probably still be gatekeepers for those films like there's gatekeepers to books but there's currently a much wider range of books than movies and I expect that gap to shrink.
Plus, there are social talents an author-director of movies has to have which would be obsolete for single creators, like a medieval bard or even Shakespeare with his theatre business compared to a modern novelist like Dostojevski. The productive talent pool may well get bigger.
A lot of my favourite books recently have been self published, so it feels like it to me. I greatly enjoyed The Martian, Worm, Mother of Learning, and Harry Potter and the Methods of Rationality. Maybe it's because they're all a bit different than what typically gets published.
I think the filter you talk about is in the wrong place if you intended it as a counterpoint. Many lone authors create books and then publishers (ideally) choose the best ones to publish. Assuming publishers choose well, no quality is lost.
With movies, most people with a great idea will never even create a movie to begin with.
Hell, even among published books many successful works were famously rejected by multiple publishers before finding a home, eg. Harry Potter, and there's a hell of a lot of published trash, so I think traditional publishers are a terrible filter!
Editors, on the other hand, are invaluable... AI shouldn't be too far off being able to edit for spelling and grammar, surely, even if editing for plot and tone will still require a human for a while?
Yes, but as those AI will all be controlled by giant woke corporations, anyone trying to make anything problematic will likely have their license to use AI stripped from them.
I don't see any obvious way for Google to stop a non-SJ Google-clone from catering to the people Google rejects. Their AI might be a few years behind and 1/10 the scale, but they'll have it.
You need actual law (or terrorism; violence, in any case) to push things beyond "separate-and-not-equal" (the current state) into outright "no access to services", because the class of SJ excommunicees is large enough to form its own economy and can't be dissuaded by status incentives.
Yeah but as original article pointed out, scale is crucial when it comes to AI. 1/10 scale AI would likely be more than a "few years behind."
I said "a few years behind *and* 1/10 the scale". I meant "behind" in algorithm design, which does matter to some extent.
I agree it wouldn't be as good, but it could still be "good enough".
There's an excellent example of this already. AI Dungeon is a text predictive choose-your-own-adventure game that uses GPT-3. Recently, OpenAI (the owners of GPT-3) forced them to implement a heavily restrictive content filters, to the chagrin of most users. Disgruntled former fans worked to build a competitor called "NovelAI" using open-source GPT clones, EleutherAI's GPT-Neo and GPT-J, which are 2.7B and 6B parameter models, compared toGPT-3's largest model is 175B parameters.
Despite having a model 1/30th of the size of AI Dungeon's, NovelAI is superior in many respects. While its raw text prediction is slightly inferior (but not by that much -- GPT-like models achieve diminishing returns fairly quickly and EleutherAI's models roundly outperform similarly-sized OpenAI models in most tasks due to far better training data), NovelAI compensates with a ridiculous amount of quality of life and customization features that, in the hands of an experienced user, can give superior results in general to AI Dungeon despite an objectively weaker AI.
So you're forecasting no more live-action actors? It'll all be CGI and animated models?
Maybe we will have the equivalent of self-publishing today, all the ebooks on Amazon etc. But an awful lot of those are absolutely crappy content, I am willing to bet that most of the "single individual makes an entire movie" will be crappy content as well (I don't just mean the writing, I mean layout and design and typos and lack of editing etc.)
So I think Hollywood will still have an advantage, even there: if everyone can make a movie without actors, the established industry can do it faster, better and cheaper, with the brand name recognition: "Do I want to watch a new movie from Paramount or from J. Random Nobody?"
I think first it will be cartoon/animation that it will be possible for one person to create with the help of software. And eventually the rendering will be indistinguishable from live actors. But probably at no time, and definitely not in the next 50 years, will there be no more live actors. It will be a both thing.
They'll still probably be big Hollywood movies and studio TV shows produced by big teams of people and actors. But they'll also be a huge amount of independent mostly solo created content (in a similar way to books are now). The best of that will get picked up by things like Netflix and distributed widely, which will infuse the wider-culture to some creative diversity.
Probably the biggest market though will be 'better remakes than Hollywood.' So just like fan fiction is big now, they'll be tons of different non-canon sequel trilogies of Star Wars or different endings to Game of Thrones, or additional stories in he Harry Potter universe.
Right, Hollywood will always have more polish. But do I want to watch the 1000th polished Marvel movie, or a somewhat flawed indie darling that is actually original? The explosion of indie web novels makes word of mouth and reviews much, much more important, but there are true gems out there to find, and I expect movies will end up in the same way, if not more so! (more so because conventionally published novels show more originality IMO than mainstream movies - precisely because of the different level of economic risk involved, I expect.)
I'm going by Sturgeon's Law: 90% of anything is crap.
Yes, there will be great indie movies. But you'll have to wade through an ocean of crap to find the few jewels.
Hollywood will give you a reliable product. It may not be original but you know what you'll get.
I'm not denying there would be great original 'made this in my bedroom' stuff, I'm saying that if every Tom, Dick and Harry is doing it, you're going to find it very, very hard to go through all that content and find the good stuff. It'll be more likely the really good ones will eventually be picked up by Hollywood - or Netflix, Amazon, HBO, etc. to produce content that can be easily mass-marketed.
You have 1 million indie shorts on your Youtube algorithm curated recommendations list. Are you going to wade through all 1 million or are you going to pick "I really like Tom Thompson's stuff" and stick with that niche of "stuff like what Tom Thompson makes"? Because I think that is how it will end up.
"I guess at the limit this is just banning insider trading, which is supposed to be good."
I think most economist think banning insider trading is bad for getting the market to reflect the best guess of where the price should be because it is leaving out a piece of true information held by the insider. So it does make the prediction market work worse.
I think the argument against insider trading is more of a principal agent problem. Should the insider be using his privileged position to be making money of the stock holders who he works for.
Re. "a few journalists who are scared people will make fun of them if they jump the gun.:" I think media bias is a more likely explanation.
The two are not mutually exclusive. "Not wanting to look foolish by calling a race too early" seems a very plausible media bias.
Likewise, I expect most news outlets to be biased in favor of their stories getting more click and views. "Too close to call! Stay tuned for continuous updates!", if it seems plausible to casual viewers, would probably get better ratings than "Welp, looks like the Republicans have this one in the bag."
I'll note that if I were pro-Democrat, and I believed the standard model that Democrat turnout is more mercurial than Republican, I'd report everything remotely marginal as "too close to call" to get out the vote.
It's not a more likely explanation. Too close to call is very common terminology that's been used by media election desks for at least the last few election cycles, independent of which party is leading.
"even if you double your money, 100% return over eighty years is a bad investment."
Honestly, this is a problem over even two years--a 10%ish opportunity cost is a lot.
But can't this be resolved by purchasing conditional shares in some portfolio where the shares are traded freely? Like an SP500 index conditional share?
So I purchase a conditional share in a "China GDP 2100 vs. US GDP 2100" SP500 index fund. The value of my share increases in lockstep with the increase in the fund, which over the course of 100 years is a fine return. Unconditional fund shares can also be bought/sold as usual. In 2101 the conditional shares will turn into real shares, or into nothing, depending on whether the condition was met or not.
This seems to solve the liquidity and returns issues both.
(Tax loss harvesting from losers might skew some numbers temporarily, but that's even a good thing in that it provides non-arbitrage liquidity)
But I'm obviously not the first person to think of this, so...why isn't it standard?
The media has a well known incentive to always describe races as closer than they actually are so that people will keep breathlessly reading updates. The same thing happened in the 2012 presidential election.
This is everyone(-ish) said about Nate Silver claiming Trump had as high as a 30% chance the weekend before the 2016 election: "He's just doing it for the clicks!"
Ironically, 2016 seems to be the one case where the media did the opposite, apparently due to sheer wishful thinking.
The biggest weakness in Forecasting today is the unknown unknown, meaning, getting the most relevant questions forecasted in the first place. For instance, it's more relevant to the world to forecast the probability a pandemic will occur in the next 3 years then to predict the level of active Covid cases in 3 months, given a known Covid pandemic. Also would have been extremely useful if forecasters had foreseen the supply bottlenecks 6 months in advance (because then they would have been less likely to have happened, given that forecasting markets had high enough visibility. No, this wouldn't have been a paradox: the odds could have risen and declined over time as the forecasting market spilled information into the market market.
Has anybody seen my black swan?
A pandemic was a black swan, fair enough. But the drop in industrial production a year ago should have given more than a few people the bright idea that: when this thing is over, demand will come roaring back and when it does inventories will be low... and there will be a crush at the ports. That's wasn't a black swan; that was foreseeable. I didn't see it, but I'm sure others did. Perhaps they kept it quietly to themselves and put their money into crude oil futures...
Yeah the trick with unexpected events is that no one in their right minds really bets on them. Hedging is different, of course. I agree about the adaptation phase. When things started to get messy, one had to be more imaginative and less risk-averse than before to forecast stuff.
"Prediction markets reached near-certainty about the winner while traditional media was still talking about how un-call-ably close it was. Apparently having hundreds of people all incentivized to give precise probability estimates very slightly earlier than the next guy, works better than having a few journalists who are scared people will make fun of them if they jump the gun."
Compare this to how PredictIt wasn't willing to call the 2020 election for Biden until long, long after it was obvious that Biden had won. I think a better model might just be that PredictIt and some similar markets tend to err in the direction of Republican candidates for contingent historical reasons.
I wonder how much that had to do with 2016 and every media pundit and pollster serenely certain Hillary would win because look at her leads in the polls, look how she is forecast to win Florida etc., and then they got a massive pie in the face on election night.
So they erred in the other direction of "well since not *every* single vote has been counted, let's not announce anyone as the definite winner".
At least one prediction market had Hillary at 89% the day before the election, too, which would have been a chance for Scott to make some of his money back on bets against extreme prediction market certainty.
https://www.marketwatch.com/story/heres-what-the-betting-markets-say-a-day-before-the-election-2016-11-07
Amplifying small changes can make the smart people lose all their money faster when markets temporarily move against them.
Also, depending on the exact mechanism for resolving each 5-year market, which you don't really explain, the resolution could be manipulated. If there is a lot of open interest on the expiring contract relative to the volume on the new contract, people can profitably manipulate the price of the new contract to manipulate the settlement of the old one.
Something sorta similar to that happened when oil futures went super negative for a day last year. A firm had contracts to buy a huge amount of oil at a fill-in-the-blank price based on the price of the futures at a certain time. So they manipulated the futures to go negative at that time, and afterwards the futures instantly bounced back up. https://www.institutionalinvestor.com/article/b1nf4dxm53536k/The-Mysterious-London-Traders-Accused-of-Manipulating-Oil-Markets-and-the-Anonymous-Hedge-Fund-Rare-Coin-Expert-and-Day-Traders-Who-Are-Fighting-Back
> Might prediction markets be lurching to a false certainty too soon? I used to think this, I tried betting on that thesis a bunch of times, and I always lost money. I guess realistically we can’t know for sure that they’re not overconfident until we’ve tested their 98% probability fifty times, but I nominate someone else to lose their money for the remaining 40-something experiments.
At the beginning of November last year, Biden was predicted to win probably but low-confidence. On the night of the third they were medium-high confidence (but less than 90%) that Trump would win. By the time they got to 95% on any question I think they were usually right, but they definitely were prematurely confident.
Sanders is vastly overrated. Unless people are taking early cheap bets to score off his small but extant cult base. Warren is immensely overrated as well. They should both be at 1, as should Clinton. AOC is slightly overrated. Buttigieg is underrated. Harris overrated and Biden underrated. The real race is in dark horses who haven't even talked about it yet. Also the PredictIt markets are notoriously poorly moderated/modified.
Are you talking about comment moderation or something else?
Regarding Hoel's prediction #14, here's my honest counterprediction:
30% nuclear war by 2030.
I'm not about to bet in a prediction market on this, and I indeed predict that prediction markets will predict against it. This is because all predictions on prediction markets are conditional on the prediction market itself continuing to exist, which is highly correlated with the outcome of some questions. But it's why I don't live in a large city any more.
On street votes, it's worth clarifying that the even most ardent advocates do not think that ‘most’ people will vote for them. Even the Strong Suburbs report linked to in the text above does not estimate an overall take-up rate of more than about 3%. The estimated price elasticities implied that that 3% would exhaust the economic potential for profitable densification. The potential to add more homes well in low-density suburbs is so great that that low take-up can mean the production of a considerably higher volume of new homes than happens under many current zoning systems. Many suburban lots are easily capable of producing 5x (sometimes 10x) the current amount of housing on them, even without going above three or four floors.
There may be an analogy with cryonics here. I think one survey showed that the people who had signed up for cryonics assigned an even lower probability to cryonics working than did the people who had not signed up. The former group had just looked at the numbers and concluded that the potential upside was greater than the downside.
I guess I would be concerned that locals would be even more anti-development than central planners because they're the ones who have to deal with the noise and traffic. It sounds like you're saying YIMBYs also expect this but think that increasing variance (by coincidence a few blocks will be mostly YIMBY or neutral) is worth it?
Yes, it's partly about variance, in two ways. (1) Some streets will be much less risk averse than others, as you say. (2) Even if homeowners are on average more anti-development than planners, then we get a double option on each potential development; either it can be allowed by the planners or by a small group of homeowners. The proposed legislation doesn't let homeowners block development that the planners allow. So it's strictly superior in terms of housing supply. And many of these suburbs produce almost no new housing at all at present, so it would be difficult to reduce that even if there were unforeseen consequences and different legislation in future.
But it is primarily driven by the details of the process, on two principal dimensions.
Number: many individual homeowners are willing to do more with their lots, including selling for redevelopment (at a potential profit of over a million dollars per lot in some areas). If they weren't, there would be less demand for zoning rules to protect other homeowners. But the incentives are far more diluted and harder to obtain when the decision requires a large number of people. It would be theoretically possible to rezone the whole of Palo Alto and make every homeowner much better off financially because of the enormous latent demand to live there, but they're likely to see it as risky and it would be incredibly difficult to get so many people to agree. With ten or twenty homeowners, the incentives are vastly stronger – nearly as strong as for a single homeowner – and the costs of persuading them much lower.
Mechanism: one-shot direct democracy as we are proposing inherently involves fewer veto points than layers of committee-style structures. There is much more likelihood of blame avoidance where there are multiple primaries and other selectorate mechanisms operating to make officials and politicians more risk averse. Persuading committees of people who have to get re-elected to take real and possibly controversial action is hard. And many zoning systems have accreted layer after layer of veto points in the process, with the intent to reduce controversial change.
We have good evidence from Tel Aviv's TAMA 38 process and from Seoul that people do vote for redevelopment through direct democracy, even with the 75% majority as required in Seoul (e.g. https://capx.co/seoul-searching-does-the-korean-capital-have-the-solution-to-the-housing-crisis/). And we know from Nolan Gray's work on Houston that many streets and blocks did not take up their right to opt out of the minimum lot amendments that allowed more housing there.
So the downsides seem minimal and if we achieved even a fraction of what was done in Tel Aviv or Seoul, it could substantially improve housing supply.
The American Planning Association (many of whose members are enormously pro-housing and are presumably familiar with their local political obstacles) thought it was interesting enough to invite us to write an article recommending people try it. https://www.planning.org/publications/document/9219200/
Erik Hoel's prediction #1 (that there will be a Martian colony in 2050) is not only embarrassingly ridiculous, it doesn't even follow his dictum of -
"If you want to predict the future accurately, you should be an incrementalist.."
An incrementalist would look at what proportion of the "getting a human being back from Mars, alive" problem has been solved in the 50 years since we put a man on the moon. And the answer is just a few %. The task, with current technology, is not remotely possible - which is why NASA talks (or fantasizes) about nuclear propulsion in space or setting up a mining industry (remotely, with robots!) on Mars to provide the fuel to power a return journey.
The irony is that after criticizing other smart people for "..imagining a sci-fi technology that doesn't exist" he then goes ahead in his very first prediction and does the same. He simply "imagines" that currently completely unsolvable problems will suddenly become easy-peasy such that there will be "a city on the red planet".
No there won't! There won't be anyone on the red planet in 2050!
I'd add that I think what Elon Musk has achieved with SpaceX over the last 20 years is nothing short of incredible. But he hasn't even begun to solve the really big conundrums that make currently getting to Mars and back (for humans) completely impossible.
Musk really hasn't broken any new ground that wasn't completed by the 1960s. He's done it cheaper, and he's done it in a way that's easier to replicate. At this point, his company is acting more like an engineer than a scientist (refining rather than inventing).
If, and that's a pretty big if, he keeps going at the current rate, there's a good chance he'll progress beyond what humans have done in the past. So far, he's mostly just treading old ground. He hasn't even done some of the major things that were done 50 years ago, like going to the moon. Or more recent items like going to Mars, or sending a probe out of the solar system.
Funnily enough I think the one major thing he has done is in the direction of making going to Mars more feasible - he almost alone in thinking that heavy lift rockets need to be re-usable. And I think the progress SpaceX has made in that direction has been truly ground-breaking. Sure, there are a dozen more major hurdles, but it more progress than anyone else has made in the last 50 years..
I think you're inferring a prerequisite for feasible return that's not there.
Do you really think that 'Die when you get there' trips to Mars are ever going to happen?
Colonization implies moving there permanently; living on Mars for decades doesn't necessarily require the ability to return but it would be uncharitable to describe such a trip as "Die when you get there."
Sure, fair enough. However the first trip is absolutely only going to be made if it includes a return. Which is another reason it isn't going to happen anytime soon. And for sure not before 2050.
You might be inferring. He’s possibly implying.
I mean, no-one was really trying to advance manned space flight between ~1970 and ~2010, so it's not that surprising that there was no progress and some tech even moved backwards. Given that Musk is willing to throw billions at getting to Mars, it's not that implausible that he might succeed - historically, the Apollo project showed how fast tech can advance with money thrown at it, and the lack of anything since then is squarely due to a lack of investment.
I'll grant you that the achievements of the Apollo project always give pause to Mars scepticism. For me, the progress From Kitty Hawk to Apollo 11 is even more arresting.
However, just throwing money at problems doesn't guarantee solving them, or solving them within the next few centuries. Nuclear propulsion in space may indeed work, but then again it may not, or at least it may take hundreds of years and trillions of dollars of investment. If not nuclear it'll be fuel produced on Mars which is Musks preferred solution, and the evidence for the feasibility of that is almost non-existent.
To take the other major impediment to getting to Mars, the record for the length of time a human being has survived in deep space - in the radiation 'death zone' is 12 days. And though I take your point that nobody has been trying to extend this for the last 50 years, we still don't have the faintest idea how people will survive 500 days of severe cosmic radiation let alone solar flares. We might find a way, but it might take (again..) centuries.
I wish Musk and NASA and anyone else who tries to get a human being to Mars and back all the luck in the world, but I'm not holding my breath.
My prediction is that there will be people attempting to colonize Mars by 2050, and most (and perhaps all) will die either there or in transit. And they will be lauded as heroes.
By comparison, consider the bottom of (Earth's) ocean. There is considerable mineral wealth there waiting to be utilized, it's orders of magnitudes closer, and orders of magnitudes less hostile to human life, but I don't hear about efforts to build colonies there. Why? I think it's all about glamour.
Scott has written in 2015 in a short fiction story about (among other things) a person who could see one month into the future:
> You resolve, on the first day of every month, to write down what you see exactly a month ahead of you. But what you will see a month ahead of you is the piece of paper on which you have written down what you see a month ahead of that. In this manner, you can relay messages back to yourself from arbitrarily far into the future – at least up until your own death.
So he has already invented iterative prediction 6 years ago!
P.S. The story is called "...and I show you how deep the rabbit hole goes"
“Too close to call” means what, < 99% confidence? Because no media outlet wants to retract a call, ever, and there’s no penalty against being late with a call.
on iterated prediction markets
I think a lot of prediction market design problems are already solved in financial markets. Suppose you have a non-dividend paying stcok (say, Google's parent Alphabet Inc.) The fundamental value of this stock is, theoretically, a function of very distant future cashflows, it is the present value of dividends that might be paid in 20-100 years from now, when that company will stop its crazy growth and will start distributing cash to shareholders. This long (and uncertain) horizon does not prevent a liquid market in Alphabet's stock. Indeed, right now this market is able to assign a very tight range of values to this present value, $2978.00/$2979.84.
How does the financial market handle the very real impediments to prediction markets that you describe?
1) Removal of intermediary credit risk. You can be right about your prediction but the probability that a centralised provider of a prediction market will fail grows quickly with time horizon. I would not bet more than even odds that a single current centralised betting market provider will survive until 2100. So one can be right about US vs Chinese GDP in 2010 and still lose 100% on this bet when market for the chained bet #6 fails in 30 years time. This problem can be resolved by DeFI ("can be", not "has been" as I am not sure we can be that certain about existing protocols.)
2) Diverse participants time horizons. For a market to provide a liquid tight price one needs to have participants with diverse time horizons. Market for Alphabet stock is provided by people who intend to hold this stock for less than a second. Plenty of traders hold this stock for less than a day. This should not be a problem in most prediction markets.
3) Availability of leverage. 100% return over 80 years is, indeed, a paltry reward for a correct long-term prediction. This does not create a problem for equity markets for two reasons - the mixed horizons mentioned above and availability of leverage. The money invested in a liquid contract is not frozen until contract maturity. One should be able to borrow against market value of this contract at pretty low interest rate. This is not something available today, but there is no reason for this to be unsolvable by DeFi.
4) Fragmented markets are less efficient. Equity market is much more transparent and efficient than bond and option markets. The main reason for that is that for each corporate there typically only one stock but dozens or hundreds of different bonds and thousands of options traded, so liquidity is spread much thinner. Chained iterated market, especially one with bins would fail on this count quite badly.
Thus, I do not see why a much simpler solution to a prediction market for US/China GDP in 2100 would not work better than an iterated market. On a trustworthy DeFi exchange, create a contract expiring in 2100 and settling at 0/100 depending on the binary outcome. If this contract becomes liquid enough, it will attract traders with diverse time horizons. It will be also relatively stable and it will be possible to borrow against it, so the limited return will not be major problem. To enable higher leverage, somebody might introduce options on this contract, may be even binary range options you are describing, but the contract can be perfectly liquid without them.
#1 requires that the oracles that the defi contract points to still exist and are still trustworthy. That's just as much of a longshot as PredictIt still existing in 2100.
#4 is a good argument against chained prediction markets
Ah ha! I wondered where all those new subscribers were coming from! Really and truly appreciate that shout-out Scott, as I'm a longtime fan. It's also insanely impactful to a smaller substack like mine - I think around a 20% increase in total subscribers in the ~24 hours since this post went up
Is this sarcasm? Or maybe you mean only in abstract theory?
> My only concern is that it doesn’t have the same sort of hits-you-between-the-eyes obviously-there’s-no-way-to-bias-this quality that prediction markets do.
I always thought prediction markets had the opposite quality: that they're super easy to bias. I admit I haven't kept up with the research on this, but prediction markets of real-world events lack of the property that people's predictions have to converge on the actual answer, because:
* people aren't really financially constrained by predicting wrongly, since they can just get money elsewhere and aren't going to go broke on bad predictions, so over time bad predictors will stay in the game
* adding more information (or getting closer to an event) doesn't necessarily make people better at predicting what the outcome will be, due to people's tendency to just.. be wrong, misled, or biased (not a property of some types of predictions, like things that have some random distribution, or where polling or something provides an increasingly accurate prediction of the outcome).
(arguably investment funds have the first property, for comparison, because they deal with quantities of money at which bankruptcy is a real risk.. but the real-life dynamic that old funds go broke and new ones are founded might mean that overall the market is full of funds that won't not survive actual market fluctuations over time; they've not had time to be 'weeded out' yet.)
Does substack really not allow editing comments? hm. I of course meant "won't survive", not "won't not survive"
> If there are a lot of low-quality forecasters in the tournament, then since high-quality forecasters will accurately predict that low-quality forecasters will give a low-quality answer, everyone will converge on the low-quality answer.
> They also admit this incentivizes teams to ignore “secret knowledge” that they have but which they expect other teams won’t. Their solution is to make teams very big and full of smart people, so that it’s unlikely other teams will miss something. I guess at the limit this is just banning insider trading, which is supposed to be good.
Isn't this what Bayesian Truth Serum is supposed to solve?
In the Erik Hoel link, the graph about single parenthood seems to be showing a different statistic than he says it shows (percentage of single adults who have dependent children versus percentage of children raised in single-parent homes; although these are correlated in practice, it would be theoretically possible for one of them to be close to 100% while the other was close to 0%; and both of these are different again from the percentage of children raised in single-parent homes, which the post also mentions). Also, the post refers to a percentage figure that isn't shown on the graph ("In Denmark, for instance, the total percent of single-parent households in 2019 was around ~33%"; even given the previous point and replacing "percent of single-parent households" with "percent of single adults who are parents", the graph shows about 20% of single women and about 8% of single men in 2019 were parents, so about 14% of single adults overall, so I don't know where 33% comes from).
If I saw this in a random place on the internet I'd just assume the author didn't understand what they were talking about (the first issue is similar to the widespread confusion between P(positive test|disease) and P(disease|positive test), and the ~33% could possibly have come about by thinking that the overall percentage comes from *adding* the male and female percentages, which would give 28%, possibly combined with some misreading of the chart axes to get it from 28% to 33%).
But since this was linked from ACX and recommended by Scott, and since none of the other commenters either here or there (who are clearly highly numerate and also nitpicky) have picked up on it, my prior for it being my misunderstanding is much higher.
I commented on the post asking where the figure came from (in case it was from somewhere other than the graph and I was just misreading the post as attributing it to the graph) and he replied saying it was from the graph, but didn't reply to my follow-up comment querying the issues with it.
It could be that there's some formula or heuristic (with, necessarily, a lot of demographic assumptions) that I'm unaware of, that estimates a likely percentage of single-parent households from a percentage of single adults who have children, and that feeding 14% into that gives 33%, but he doesn't mention anything like that at all.
Can anyone help shed any light on this?
I wrote the Street Votes question. I also wrote this question recently on new top charities for GiveWell. (https://www.metaculus.com/questions/8408/givewell-recommend-breastfeeding-promotion/)
If you'd like to pay me to write* questions for your organisation then DM me on twitter (https://twitter.com/NathanpmYoung)
*I can only write questions, I won't approve my own - that has to be done by other moderators.
@Scott As I've said elsewhere, if you want questions writing, please ask. You can have them on the house.
Re the election results. Journalists, particularly 25 hour news channels like CNN have an incentive to frame everything as close as possible because if they say the result the obvious people won't tune into their live coverage. This is fairly easy to do because both campaigns will also say it's close because that encourages their voters to turn out. Nate Silver has talked about this a lot in 538 podcast and his other writing.
In "bad uses of prediction markets", I present https://twitter.com/PolymarketHQ/status/1461461169223393288
Yup, that's a prediction market on whether other crypto investors will buy something. And the exchange is promoting this on Twitter with Pepe the Frog memes and suggesting you missed out on 7x profits.
I don't see how anybody could think that will provide useful information.
I wrote the street votes question. Some comments
"This is near the top of the YIMBY movement’s policy wishlist - they seem to think most people would vote for denser zoning, though I have trouble understanding their optimism."
I don't think Street Votes is near the top of the wishlist. I think that's stuff like removing zoning. But it is near the top of what seems able to pass, especially in the UK. It's not that this will make most people vote for more density, but that it's a lot better than the status quo, where there are far more veto points. I think (80%) that in Israel street votes lead to significantly more housing, largely because it allows citizens to make a load of money having their whole street upgraded at once.
tl;dr It isn't great but it's better.