Here's my dumb prediction: Neither Trump, any member of his family, not any protege closely and personally linked to Trump will be a serious candidate in 2024. 75%
Because each day in coronavirus lasts 100 years, this may not seem true, but Trump was POTUS 34 days ago. His relevancy will fade a lot in the next several years, and people who currently claim that he's the best ever will quietly shift their opinions about him.
To me, "not a serious candidate" means that the person won't be on the GOP ticket, and if they run 3rd party they do not break 4%.
If an immediate family member of Trump runs, they'll break 4% without even campaigning. The Cult 45 crowd will get them that far on unquestioning loyalty alone. Even four years from now.
Maybe! But I think people underestimate the strength of the two-party system, and overestimate people's current protestations of loyalty.
I mean, my prediction is 75%, not 99%. That's because of two things:
1. I do think that Trump has an unusual amount of personal cred with some elements of the GOP.
2. Also, the media on all sides is addicted to talking about Trump and extending his relevance.
Still, I overall think that a huge percentage of the people who right now are like, "OMG, TRUMP" are gonna be, in 2023, "Yeah, I was always kind of dubious about that guy."
I tend to agree with you. The political landscape 4 years from now is very hard to predict. Imagine calling that Trump would win the presidency in 2012, or that Obama would win the presidency in 2004. That would be foresight at a deific level.
In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies. This is very different, but much more attractive, than Trump's previous disconnected constellation of positive self-image affirming tendencies.
It's also worth pointing out that the potential for wit of a given candidate is heavily dependent on who is advising them. Observe Trump's 2016 campaign with his 2020 one. His advisers back then were of much higher quality, and I think that's in part because of the transient power he appeared to offer.
What I mean by this is that, keep an eye on political players that drift just below the surface of figure heads like presidential candidates.
>In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies. This is very different, but much more attractive, than Trump's previous disconnected constellation of positive self-image affirming tendencies.
I'm not sure it is more attractive to a large proportion of the base who supported Trump, because I don't think a cohesive set of policies was a criterion they paid much attention to in the first place. If anything, it might have ended up being a burden to his popularity, if it gave too clear a way to judge how much of his platform he actually managed to accomplish.
I think there has been a tremendous overreaction to Trump's win and subsequent predictions of what the next surprise success story is going to be. Trump barely won a clown car nomination process with a minority of support and the open disdain of much of the party and conservative media apparatus, which then immediately fell in line and started telling themselves stories about why he appealed to them when the only reason he appealed to them was he won and had an R next to his name. The sheer level of partisanship in the country is going to lead to wild swings in party profiles as voters and pundits try to convince themselves they have actual reasons to support whoever gets lucky at the right time in crowded and long nomination battles. It probably won't lead to much in the way of policy or governance differences just due to the intransigence of Congress and the judiciary, but it'll make the characters of each party seem like they've shifted a whole lot between 4-8 year periods.
Plus, for the people who legitimately loved Trump from the beginning and didn't just change their minds once he won the nomination, too much of his appeal is tied up in his unique personality and celebrity. He's a billionaire who married a model and fucks porn stars on the side. There's a "man, I wish I could live my life like that" quality that boring nerds like Tom Cotton and Josh Hawley can't replicate even if they're more ideologically coherent.
"In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies."
This is as likely as Forum for Democracy getting anywhere. Which is to say, it isn't likely. Trump didn't win because he was a "populist", whatever that means, anyway, he won because he was a rich businessman, which caused the low IQ to think he was good on the economy.
The linked Metaculus prediction was 15%. Your 75% prediction that Trump or a family member doesn't break 4% is compatible with a 15% prediction that they or someone else breaks 5% as a third party candidate.
I was basically trying to differentiate between "unexceptional third party candidate support" and "exceptional (but still doomed) third party candidate support." Without looking carefully, my impression is that in a typical year, there are third party candidates who get about 3% of the popular vote, so more than 4% is unusually high.
I agree with this, but for different reasons. I think hard as this may be to believe that it's Trump himself who won't be interested in running. Indeed, I think he wasn't interested in being President by the end of his term. His entire trajectory from about late October to the end was *not* that of someone who is (1) fairly intelligent and (2) genuinely interested in maintaining political power. I don't think he's stupid.
So why did he *say* he wanted to keep power? We can't fully discount his personal enjoyment of kicking up a fuss, because he does like that. But my suspicion is it was all about the money. He is under very serious financial strain, and by provoking a crisis and selling his true believers on the urgent, urgent necessity to go all out to respond he raked in something like $250 million between election day and the Inauguration. That will go a very long way to easing his immediate personal financial difficulties.
So while I don't think we should discount the possibility that he appears in various places to fling grit in the gears, because he likes that, I think he will find some reason (or engineer it) such that he doesn't personally run ever again. I think it's kind of been there done that, for Trump personally.
Scott, didn't you link to papers showing India had almost as many people infected by now as the entire U.S. population? India obviously has had more COVID deaths than the U.S.; official deaths seem underestimated (conservatively) by 10x. So this must be wrong:
"I haven't seen these numbers, but I haven't seen anyone make a plausible case some other country is off by a factor of two. This was mostly in there to cover China (or someone) lying about their death count in an obvious way, which didn't happen. True."
Indonesia might also have more COVID deaths than the U.S.; its official figures are ridiculously underestimated.
That was in August. Months pass by. Let's go by that paper's estimate that true infections were 26-32x official cases (and there's no reason to think that ratio has changed). Official cases on Dec. 31 were 10,266,674. This implies between 267 million and 328 million true infections and between 1.3 million and 3.3 million true deaths. This is surely much higher than the American total. This is consistent with a later seroprevalence survey between December 17 and January 8 which reported 21.4% of adults surveyed had been infected with coronavirus, which implies total infections (assuming the same % children were infected, though it was reported higher) of 292 million infected, which would imply between and 1.5 and 2.9 million deaths by December 31. That's a lot more than in America at the time.
Where do you get your IFR from? Why do you think the real death rate is 10x the reported, let alone "conservatively"? Your first link says that based on age distribution, you should only expect 3-4x, so a death toll equal to or less than the American. There's a lot of uncertainty, so it could easily be over, but your own source says it's under.
Even ignoring risk premiums the efficient market hypotheses doesn't predict a 50% chance that one asset will outperform another. It predicts that any two assets will have the same *expected* returns. But if one of the two assets has an asymmetric risk profile (for example a large probability of doing worse balanced out by a tiny probability of doing *much* better), the chance of doing better might not be 50%.
Yeah, I was going to point this out. I wonder which direction the asymmetry goes in the case of Bitcoin. At this point I wouldn't be shocked if it rose meteorically or crashed horribly (or both).
Re question 11: Hdroxychloroquine is not a particularly effective treatment for covid. It does, however, seem to be remarkably effective as a prophylaxis. This is likely how India beat covid against all odds. India accounts for three-quarters of the malaria in Asia and large swaths of its population take hydroxychloroquine regularly because it is less expensive than more recently developed antimalarials.
You are the physician, Scott. I'm just an engineer. But my understanding is that India has beaten covid when we expected it to be devastated. Some say it is masks, but they don't wear masks. Some say it is genetic immunity with no evidence. Some say it's just because most of India has no sanitation and therefore robust response to infection. No one credits the widespread use of HCQ there. Why? Only because Trump recommended it. I am honored that you actually read my comment.
I would also be shocked if any significant number of Indians used hydroxychloroquine regularly. Just because you live in a malarial region doesn't mean you take lifetime malaria prophylaxis pills.
I won't quibble, but I did read a statistic to that effect, although perhaps not on a reliable source. But the effect is real. So, maybe its due to the masks they don't wear.
You have long been my hero, Scott, and I don't want to argue with you about things of no consequence.
It was my experience, visiting rural India recently, that most of the people there cannot afford masks. Literally. It looked to me as though maybe 1% of the people around Bangalore were wearing masks. And only 20% had eaten, that day.
Malaria is not widespread anymore in India. I visit often. The travel clinics in the U.S now say there's no need to take HCQ once a week prophylactically, unless going to specific parts of specific Indian cities. India might have beaten malaria in many cities, using DDT.
And mask wearing has been fantastic in big cities in India.
Covid numbers are edging back up slowly there again :( now.
"But my understanding is that India has beaten covid when we expected it to be devastated."
The Indian COVID pandemic was basically the March-May NYC outbreak stretched out over four times as long a time. Official Indian statistics dramatically understate cases and deaths, and I expect that to be the case in Bangladesh and Pakistan as well.
I've been using India as a baseline for judging the US.
Basically, it has like 4x the population, densely packed cities, widespread poverty, lots of international travel, but not a particularly powerful central government (unlike China). We should naively expect India to have a worse time than America. Since India has 150k deaths compared to America's 500k, we can estimate that at least several hundred thousand US deaths are due to America-specific failures (as opposed to "too much rural population" or "price of democracy" or something).
I know this is a massive oversimplification, but it sounds like you have a better understanding of the Indian situation. How much might India's death count be off by? Is there an obvious reason their death count is lower that can't be attributed to a failure of US government/civil society?
People in India get the BCG vaccine for TB. If there is a biological rather than social explanations for less COVID deaths in India it is probably permanent priming of the immune system by BCG. BCG results in a lot of interferon production which might enhance both innate and specific anti-viral responses.
I don't think it's necessary to conclude that you and Bucky were "worse" than guessing 50-50 on everything. For example, such guesses would imply a distribution with 0% of probability mass between 100k and 3M US COVID deaths. The predictions look pretty reasonable to me; I know that a lot of people (myself included) were underconfident about the possibility of a 2nd wave.
Also, isn’t it weird if the scoring rule doesn’t penalize you for answering 50% on three questions about the same number at three different levels, separated by orders of magnitude?
Is it possible for the real probability of X > Y to be 50% at the same time for Y = 100k and Y = 1M? I guess we could be in a world where future X is either 1E9 or 1E3, with 50% probability, in which case the two answers are both perfectly fine. But it still feels as if there should be a penalty there in the real world. What am I missing?
Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths. There's almost certainly a lot of undercounting going on (though it may not be deliberate lying). The UK, with their unusually loose standard for a death within 28 days of a positive test being a Covid death, is one of the only nations where COVID deaths and excess mortality are similar - in fact the UK has slightly more COVID deaths than excess mortality.
That still isn't enough to knock the US off the top afaiw though. But of course, China's figures are worth as much as a bet on Trump being president, but I suppose it doesn't qualify for being obvious and no one will ever know the true death toll.
> Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths.
This assumes that the only difference is covid. We know that isn't true; the societal reaction had wide-ranging effects.
> Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths.
Not necessarily. Some of that may be deaths caused by the reaction to covid; I've at least read articles about suicides being significantly up, and people being reluctant to visit the hospital if something might or might not be wrong (which a lot of people were when the panic was at its worst - everyone knew the hospitals had covid patients) is going to give you some missed heart attacks and strokes. On the other hand, traffic deaths should probably be down - there's much less driving, even if what there is is sometimes more reckless. Last I heard we had anecdotal accounts but not a lot of actual data on the non-covid breakdown of the excess deaths, but that was some time ago and I would love to be proven wrong.
This might be true for some countries, I guess. But at least for the Czech Republic (where I’m from), our terrible total excess deaths are totally correlated (0.94) with our terrible covid infections, so no extra deaths to be seen. (Yet! unfortunately, reduced prevention will probably result in excess deaths in future.) For a chart, see e.g. https://boundedlyrational.substack.com/p/nadmrt-a-kauzalita
"That still isn't enough to knock the US off the top afaiw though."
It is. India has almost certainly had more COVID deaths than the U.S., and there is a significant chance (30%) Indonesia has had more. Russia is coming pretty close to surpassing the U.S., as well.
Yeah; it is true. Official cases in India have been fairly consistently about 1/30 of infections. Even assuming the IFR in India is half that in the U.S., it still has a lot more deaths than the U.S.
Not sure why that assumed IFR is supposed to be reasonable. I agree that total infection counts would be a lot higher than reported but it seems very likely that serious/critical cases in India were still a smaller fraction than in the US. Across south Asian countries, deaths per million population are 10-20 times smaller than the US or the UK. I would assume some undercounting, but while it's easy to miss infections, hiding that many deaths (inadvertently or on purpose) doesn't seem probable. Based on (admittedly spotty) reporting, I would assume an upper bound of ~(2 x reported) on actual deaths in India. Here is one example of excess death reporting: https://science.thewire.in/the-sciences/covid-19-mumbai-all-cause-mortality-data-ifr-bmc-seroprevalence-survey/. It's not representative of the whole country but Mumbai had one of the (if not the) worst outbreaks in the country, which disproportionately hit slum areas (so probably should have a higher fatality rate than average).
"Across south Asian countries, deaths per million population are 10-20 times smaller than the US or the UK."
There is no reason whatsoever to believe that's actually the case.
"Not sure why that assumed IFR is supposed to be reasonable."
There is absolutely no chance Peru's and Mexico's IFR was lower than .5%. Peru's and Mexico's median age is very close to India's. Assuming the seroprevalence surveys in India are even close to accurate (and there is every reason to believe they are), India has had at least twice as many COVID deaths than the United States, and likely more than 3x as many.
Also, Mumbai is one of India's richest cities. In Russia, where deaths are understated by a factor of 6.3, they are understated by only about a factor of 2 in St. Petersburg and Moscow. I am confident that the reporting divide between core cities and the other parts of the country is even worse in India than it is in Russia, and India's core cities make up a much smaller portion of the population than Russia's do: https://raw.githubusercontent.com/dkobak/excess-mortality/main/img/regions.png
I am not convinced on the utility of comparing IFRs. It's hard to trust infection counts even in the US, where there has been a lot more RT-PCR testing than any of the other countries you mention. IFRs are normalized by total infections but that denominator does not reflect actual infections, and it's hard to say how off it might be in different countries. How do you account for the variable testing and reporting issues across countries?
Excess mortality reporting, although patchy right now, is more useful for bounding the total deaths. Several Indian cities and states conducted death audits to account for excess deaths and it shows undercounting of "COVID deaths" but nothing supports even a factor of 5 undercounting.
The only link from you in this subthread is about excess mortality in Russia. You've maid claims about India (such as that official cases are 1/30 of infections, and claims about the number of deaths based on the number of infections), Indonesia, Mexico, Peru.
I have a pretty strong prior that any cause of death that disproportionately impacts old people will be undercounted, not out of malice, but just because we don't generally investigate too hard on cause of death when someone dies in their 70s or 80s, so unless they had a diagnosis prior to dying, it won't get counted. This means nothing for international comparison because all countries would have the same issue, but I am consistently surprised that people think Covid deaths are overcounted.
I'm almost certain there has never been a consistent nationwide policy on that type of thing. It's up to county level medical examiners to deal with that.
"Will the Bay Area stay locked down until Election Day (11/3/2020)?" was worded in your original post as "Bay Area lockdown (eg restaurants closed) will be extended… until Election Day". Both wordings to me are unambiguous: they imply a single continuous lockdown. If they meant multiple lockdowns, the former would read "Will the Bay Area be locked down on Election Day" and the latter would read "Bay Area lockdowns will continue on and off".
Zvi also explicitly clarified on his post (https://thezvi.wordpress.com/2020/05/01/slatestarcodex-2020-predictions-buy-sell-hold/): "If anything, that seems high, assuming it means continuous lockdown until then rather than being locked down on election day." I think omitting that result from his score would be better than including it, because he was explicitly predicting a different thing from you.
I don't think restaurants ever opened for normal business before Election Day. I realize this was ambiguous, but I was trying to get at the idea of "when will the pandemic be over".
Incidentally, I've really wished that no one would use the word "lockdown". It seems to mean totally different things to different people. My friends in France were telling me a few weeks ago that restaurants and shops are closed, and you're not allowed to travel more than 100 km from your house, but the government is saying there's still a possibility that they will go back into lockdown. Meanwhile, in Texas, people think we're still in lockdown because you're required to wear a mask when you walk into a restaurant until you sit down.
I'm getting different totals for the scoring. Could you/someone else check to see what's going on? Just to make sure we're doing the same calculation: Let's call p the probability that a forecaster assigned to the correct answer, then LogOdds = log(p/(1-p)).
Some examples: scott's score on the first question was log(0.6/0.4) ~= 0.405. His score on the third question was log(0.9/0.1) ~= 2.197 (balanced out exactly by his score on the second question).
Doing this computation for all questions gives total scores for Scott, Zvi, and Bucky at 7.2034, 13.2863, and 6.5006 respectively, all better than chance!
Also, I recommend switching to Brier score (squared error) for things like these. It's a proper scoring rule, which means that you cannot game the system; in expectation you'll get the best possible score by reporting your true beliefs. In contrast, something like LogOdds can be gamed since increasing the extremity of your forecast just ups the ante, but the odds of the bet don't change [i.e. log(p/(1-p) = -log((1-p)/p)]. For instance, if you think something is 90% to occur, then by reporting your true beliefs you are expecting to get 1.7578 "LogOdds points" (0.9 * log(0.9/0.1) + 0.1*log(0.1/0.9) ). However, you could expect to score more points by reporting a more extreme belief, like 95%. In that case, you would expect to score 2.3556 points (0.9 * log(0.95/0.05) + 0.1*log(0.05/0.95)). Doing the same calculation with brier score for belief p and *reported* belief x has you minimizing the function p*(1-x)^2 + (1-p)*x^2 for values of x, which occurs only when p=x (note that lower Brier scores are better).
Here's some Matlab code to run these computations:
He used the logarithmic scoring rule for the calculations (calling it "log odds" in the post is an error). The score on a question is just ln(p) - ln(.5). This is a proper scoring rule.
When I do that calculation I get a total of -0.54 for Scott, 0.81 for Zvi, and -1.95 for Bucky, which is identical to what Scott reports for two of the three but slightly different for Zvi's score.
Yeah, when I use logarithmic scoring I get the same results as you do - so Scott's calculation for Zvi is slightly off and his description is also a little wrong. Thanks!
Forgive my ignorance for question 12, "Will I personally will get coronavirus": isn't a 30% probability of this being true equivalent to 70% probability of it being false... i.e. Scott correctly predicted he wouldn't get covid?
Questions marked red are resolved as "false", black ones as "true". Predictions are then scored based on whether they're high probabilities (if the question is black / true) or low probabilities (if the question is red / false).
Scott predicted he wouldn't get Covid with a probability of 70%. The logarithmic scoring rule then says that Scott does deserve credit for this prediction (which contributes a positive term to his overall score), but somewhat less than Zvi (80%) and Bucky (90%).
Next time perhaps include a competent astrologer who also uses ML. Maybe a female for stronger intuition. It would be v interesting to see the results.
Professional gamblers are quite well attested, see eg the final table at the WSOP often having the same set of people. I can't tell if this series of posts is a troll, though.
So are astrologers who work with AI to test their predictions. I assume you know a lot about how astrology works to be so dismissive. Otherwise we have the same issue as with UFOs...people closed minded (i talk about astrology so i must be a troll) which perpetuates prejudice and ultimately precludes brilliant people from studying it and possibly open up new avenues.
I will bet you $20 that you can't find 2 astrologers, with no personal or professional connections to each other, other than being professional astrologers, whose horoscopes will agree given the same anonymous person's star chart, and in the event you do find such a pair, I'll pay an additional $10 if they agree on whether Ophiuchus should be part of the zodiac or not. As for UFOs, there's a huge difference between "it's incredibly unlikely given the size of the universe that we are the only intelligent life in the universe" and https://www.amazon.com/Space-Raptor-Invasion-Chuck-Tingle-ebook/dp/B00S4B95RQ
This makes no sense. Literarily none. Do you know anything at all about astrology? "that you can't find 2 astrologers, with no personal or professional connections to each other, other than being professional astrologers, whose horoscopes will agree given the same anonymous person's star chart" I am a western astrologer. No one I know cares about Ophiuchus as constellations have no bearing on our work other than the origins of the historical names for signs. We work with the tropical calendar which is not anymore in sync with the current position of constellations.
I talked to a few professional gamblers, and the explanations of how they worked made perfect mathematical sense.
One of them was a cab driver in Las Vegas, who used to make decent money playing card games at Las Vegas casinos, until a bunch of (or all?) casinos banned him for card-counting. Apparently card-counting gives you enough of an advantage that no casinos want to deal with you if you're doing that. If he was less obvious about it, perhaps he would have still been earning an income by gambling.
Another was a poker player in an online casino. He explained that, first, he is a lot better than the average player in room, and better enough that it's worth playing (it being mostly the other people who absorb the average loss); and second, the online casino paid people like him to fill out tables that did not have enough players.
Gladly. Please check the following two publications demonstrating that planetary positions correlate with phenomena with high r2 values written by Renay Oshop my project partner.
"Astronomy of the day is more effective than seasonal decomposition in modeling and predicting the rates of Amazon review misspellings," International Journal of Scientific Research and Management, Feb 2019, Vol. 7, 2, DOI: 10.18535/ijsrm/v7i2.aa01, https://ijsrm.in/index.php/ijsrm/article/view/2040/1719
But not to professional gamblers? I am a bit stunned by the amount of hostility here. Is this necessary? Let me explain my position. I am aware of the risk Scott Alexander took by addressing me and am grateful for his open mind. Other well-known cognitive scientists and AI researchers did the same on Twitter and Reddit but I am not going to name them to preserve their professional reputation. Look at what I am saying while science claims to have an open mind (?). This is why using ML is a good idea: to put astrological predictions on a relatively unbiased platform with large public data sets.
I think astrology should be put to the same standards as the social sciences. No more no less. Same predictive values. We are dealing with human beings here and predictions have thousands of variables; in addition, we astrologers don't have a unified body of ever-increasing knowledge behind us because of the disruption of the Enlightenment. Since then astrologers needed to work outside of the Academy. Now the Academy is in shambles and because of the Internet and an increasingly skills-based job market we are returning to a more personal way of accessing knowledge. Perhaps our time has come.
With help of ML we can put thousands of years of multicontinental astrological traditions to test. Yes, there are charlatans...there are none in science? Yes, some of our predictions were incorrect...none turned out to be so in science? But as far as I am concerned there is absolutely no competition here. I love science but not scientism. Very grateful to be alive in these exciting times.
In my view, the only two aims of astrology is to help clients overcome their challenges and make useful predictions about the world around us. I have addressed the second above. As for the first claim, I believe that we are all born with a horoscope which is a map of the planets at the time of our birth. This map is a symbolic representation: a blueprint or a bit like a negative of a photo. You can develop this negative well or badly but you cannot change the negative. A personal consultation with an astrologer is similar to a session with an analyst helping to develop the potential of the client --embedded in that horoscope to the maximum.
This is my personal view: a rather fatalistic approach giving very significant weight to the givens in the nature/nurture balance and quite aligned with our contemporary views on the role of genetics in human development.
Finally, for fun, and this is purely speculative if there is any validity to the simulation theory and in the unlikely event that we ever discovered that it would not surprise me to find horoscopes to be part of the digital construct.
I am happy to offer a private consultation online to any one of you to decide for yourselves. My only request is that you provide your correct birth data. I don't need to know your name and prefer not to see your face. Of course, if you find the consultation useless you are welcome to a full refund and publishing your opinion in any forum you desire. Jutkasastrologyandscience@gmail.com
Agreed. I do have to point something out, though. Lets assume that you're right and there is astrological research that makes valuable predictions. But then you included two examples that are really underwhelming because the claimed results do not seem all that useful. If that's really the best you have, that's not much.
You are hoping astrology can be held to the same standards as social sciences. I'm not in social sciences, so maybe the similarity is closer than I think, but it's really hard for me to see the value of predicting typos in Amazon reviews or number of celebrities' Twitter followers. If I was tasked with reviewing a technical article that predicted these things based on something I believed could predict them (such as native language, geographical location, education level, type of device used for input), I personally would not recommend publication, because I would fail to see how researching this has value. I'd be surprised that someone funded it (or not, there's been a lot of funding for all kinds of useless or ridiculous projects).
If you want astrology to be held to the same standards as social sciences, you're going to need stronger examples than what you cited.
Here is another open access astrology-based ML predictor for alcoholism vs teetotalism included below. Please feel free to try it.
Meanwhile, the examples above are not trivial at all when considering the following:
Trying to publish on the subject of astrology in a scientific journal is quasi-impossible. The prejudice is insurmountable. These particular studies were done because the subjects are familiar to the public and the author needed a huge and openly accessible free data base. Amazon provides this from the time the first reviews were written. Ms. Oshop also needed the kind of results that are not subject to interpretation, in plain words, results that are not easy to explain away. Finally, for most astrological predictors to be precise you need the exact time of birth (not just day). In the case of celebrities, they are relatively easy to attain and have the predictions publicly verified.
I agree with you about the very relative importance of Twitter followers, however, to be honest for a moment here, is this the general impression you get of AI researchers, cognitive scientists, engineers, investors, science-podcasters, etc. working either in or outside of Silicon Valley?
Admittedly, I myself have probably chosen a different subject but I am not a programmer and understand that a lot of thinking goes into how to formulate a study that fulfills all requirements on such a taboo subject. Maybe there was also a consideration of public appeal.
However, ultimately, I don't think it is relevant whether the original study is about Amazon misspellings or Twitter followers as long as it demonstrates that astronomy of day predicts with stronger values than other traditional methods. (The Amazon study does point to a very strong correlation with Mercury retrogrades, for example). Once it works you can use it for anything you like including all kinds of useful Covid related predictions.
The hard part is to have someone look at it, have the courage to say maybe there is something here (if there is). I would be so grateful if an expert looked at the study on its own merit (methodology, results, etc.) because I am not a programmer myself but then we would get somewhere.
OK,here is the other example you can try for yourself. It predicts for alcoholism vs teetotalism.
It's open access and was written for people with no knowledge of programming. You can immediately see all the problems I have described above hence I have not included it in my previous post. Also the data base is way too small which is why I would be grateful if anyone tried it for themselves using exact birth data of people you know well and therefore can easily verify the results. An acquaintance who has been betting on horses using ML for years just repeated it with his own spreadsheet and found it reliable.
"Future misspelling rates in Amazon reviews were successfully predicted using only basic astronomy data in a neural network".
I imagine misspellings correlate very highly with the number of reviews written by non-native English speakers. I'm also quite sure that the number of non-native English speakers posting reviews online depends on the time of the day. It is a lot more probable that you write a review at 4:00 PM rather than 4:00 AM. It would be trivial for a ML algorithm (or a classical algorithm) to calculate which countries were awake given so much data about positions of different planets relative to us. So it seems to me that we are looking at an fancy "what time is it" calculator. Yet there is so much I don't understand because of my lack of education that I may be completely misreading it.
I still think this way of giving and judging predictions, using confidence levels, is fundamentally flawed.
Here, let me make some predictions with confidence levels:
1) The sun will rise in the east on February 24th (95%)
2) The sun will rise in the east on February 25th (95%)
3) The sun will rise in the east on February 26th (95%)
4) The sun will rise in the west on February 27th (95%)
5) The sun will rise in the east on February 28th (95%)
6) The sun will rise in the east on March 1st (95%)
...
20) The sun will rise in the east on March 15th (95%)
If we evaluate my predictions a month from now, I think you will find that out of 20 predictions I made with 95% confidence, I got 19 correct. Perfectly calibrated! Clearly I'm the greatest predictor on earth!
I think the only good way to score predictions is to compare them to other people. If a hundred people predict A, you predict B, and you're right, that's great. If 80 people predict X, 20 predict Y and you also predict Y, that's pretty good. If 5 people predict N and you and 95 others predict M, then that's not impressive at all, even if you're right.
"I think the only good way to score predictions is to compare them to other people."
But isn't that what he did in some form or fashion here? He even says that it's not so much about absolute score as relative position, "Wringing an absolute judgment from a scoring rule like this is hard, but getting relative standings is more straightforward. I did better than Bucky, and Zvi did better than me."
Yeah my point was more a general point about the way predictions are often treated on this blog and the wider community. Something I've been meaning to say for a while but never got around to.
What Scott is doing here is a step up from how he did it in previous years. But 3 is a very small sample size, and he scores based on self-proclaimed confidence-intervals instead of based on 'distance from the norm'. Zvi scores more points for his mail-in ballots prediction than his China prediction, despite the latter being, to me, more impressive. Predicting someone that everybody else is also predicting is not impressive.
That only makes you look like a good predictor because you're gaming the system, picking a bunch of guaranteed events and deciding what rate you want to get them right. It is really obvious that you're doing this because we all know the sun is definitely going to come up tomorrow, so no one is going to think you're a great predictor. And if you can get perfect calibration while listing events that everyone doesn't think are obviously guaranteed, then it sounds like you're making valuable predictions.
Typically, the way predictive model scoring is done in industry is to compare against a null predictor of some sort. For the sun coming up, the null predictor would be yes every day for the next five billion years, so if you predict yes, your accuracy premium above the null predictor is nothing.
Scott is doing something *sort of* like that by setting 50/50 picks to be scored at 0 if they end up being true, but that isn't quite the right way to do it. I'm not totally sure how you define a baseline predictor for arbitrary binary events, though. Random prediction would work if you were actually making binary predictions, but throwing in the confidence estimations messes things up, and it only really works (and justifies 50/50 being 0) if these are truly toss ups.
But to better demonstrate why I think what Scott is doing is degenerate (agreeing with you, as it stands), take something like sports. The top team in the NBA right now has a 80% winning percentage. If I just predict they win every remaining game this season and say I'm 80% confident of that, I'm likely to end up scoring very high by Scott's scoring system, yet I have effectively just made a null prediction and should score 0.
That said, he's clearly doing the right thing with the over/under type bets, though I think in that case you have to weight the scoring by how close you ended up being and not just all or nothing.
The absolute score is not very meaningful, for exactly the reason your example illustrates: it doesn't have any way of accounting for the "difficulty" of the predictions. The point of the scoring rule is that if you are trying to maximize your score, you will use your true beliefs.
Are you suggesting that he shouldn't use confidence levels at all - that he should make deterministic predictions?
I mean, in terms of how useful someone's predictions are, I can agree a term for surety is useful - obviously, given someone whose predictions are well-calibrated, the surer they are of things within that constraint the better. However, I think somebody who's perfectly-calibrated but whose predictions are evenly distributed between 50% and 100% is probably still more useful than someone who always predicts deterministically but gets it wrong 5% of the time - if nothing else, you know which of their predictions you can rely on (if those predictions are all minor then that's its own problem, but that's more of an argument for some sort of weighting based on importance along with the aforementioned surety term than for comparing them to others).
Also, how would you score wrong predictions in your model? If I predict something bizarre that nobody else did and I'm wrong, is that better or worse than if I predict the same wrong thing everyone else did? (My understanding is that strategy-invariance would require "worse", but hey, it's your model.)
About the Alcubierre drive solution, I think people are underestimating just how low physicists' priors are on things like superluminal signaling. Describing violating energy conditions as a "technical challenge", like the metaculus question does, doesn't really capture the problem there. Also, any sort of superluminal travel that works at all like you'd want it to (by connecting two distant points in otherwise flat spacetime) necessarily implies time travel. If I had to guess, phrasing the question like "will a working time machine be demonstrated before 2100" might elicit a lower prediction than the question as it's worded now.
I never understood the argument for why FTL necessarily implies time travel. Suppose I somehow invent a teleporter that lets you step into it, disappear in a puff of smoke and appear at the target destination t seconds later where t is the distance traveled divided by ten times the speed of light. I cannot use this machine to accomplish anything a normal person would recognize as time travel, no killing Hitler or shaking hands with my future self. Why is this a problem?
Consider me unconvinced. I've parsed quickly the Wikipedia page describing how FTL signaling could lead two people moving relative to each other to hear a response to their signal before or was emitted, and it seems to rely on using relativity formulation of Lorentz transform, which FTL itself makes wrong. FTL would mean you can mesure absolute velocity, i. e there is a prefered frame of reference (for FTL stuff, not for the non -FTL physics). So no, FTL will not lead to time travel in the classical sense. It will just break special relativity and reintroduce the aether. No big deal, except no FTL signal has ever been observed
Oups, i read Wikipedia to fast and missed the last sentence : they explicitly mention that if you allow for a preferred frame (for FTL) causality is not violated. So i was right, FTL does not violate causality, it re-introduce the aether...and I stand it is no big deal, cause the insistence on being unable to define a special frame of reference is imho the weak point of special relativity. Background radiation already put it in jeopardy...
Aether is a sufficient but not necessary chronology protection for FTL. FTL can be consistent with causality without a special frame as long as there's some physical mechanism that breaks the FTL if a causal loop threatens to form.
(I've seen people propose that quantum vacuum effects would collapse any wormhole or series of wormholes at the instant they permitted a closed causal loop. I'm not sure whether a similar mechanism has been proposed for the Alcubierre metric.)
Note that I'm not saying FTL is necessarily possible (or, for that matter, that causality is necessarily true; time travel doesn't *have* to result in paradoxes, although consistency protection without causality gets bizarre fast).
Sorry, my answer is wrong (there's some wiggle room with the exact definitions of "t seconds later" though). For a reasonable definition of faster-than-light, it's more complicated than warping out and then flying back.
If your definition of warp speed is invariant for moving observers, then there will be a way to travel back in time: You'd need to be able to warp twice, and the second time, you need sufficient velocity relative to your original position. The trick is that faster-than-light travel for one observer will appear to be backwards-in-time travel for another observer with high relative velocity.
You can get around this by saying that warp speed is always defined in terms of Earth's frame of reference. You lose invariance of physics (and therefore just about all our theories of fundamental physics!), but you can regain causality.
I would not choose earth frame, but some absolute frame. Exactly back to the aether, the version whose Lorentz transformation tried to reconcile with Michelson&Morley experiment. Personally I like this framework, just before Einstein went a step further: as no mechanical nor electromagnetic phenomenon could measure motion against the fixed frame, why not get rid of the fixed frame and use a formulation that make it dissapear from the mathematical desciption? It was a good idea, as invariance also holds for non-electromagnetic phenomenons...
However, any FTL information transfer would mean that relativity is not absolute (for this you need a single max velocity), so back to a fixed frame (that you can even define using your FTL). I don't think we will see FTL, not because it would create time-travel paradox, but because c is really the max speed for anything we have observed until now. We have observed some effects very different from mechanics and electromagnetism that produced special relativity: gravity waves, electroweak, nuclear....so FTL would be a new new thing, and I find it unlikely that new new thing break an invariance that was verified by multiple new things.
Still, I am still fond of pre-einstein absolute frame: maybe because I make less mistakes using frame of alice, frame of bob, and an absolute frame with an absolute time than just alice and bob, maybe because doppler shift of cosmic background radiation looks eerily close to a physical absolute frame...
FTL travel like you describe would make shaking your own hand possible. The idea is that which events can be said to occur "t seconds later" actually depends on your velocity. Say you and your target location start off at rest relative to each other. According to special relativity, changing your velocity so that you are now moving *away* from your target effectively pushes your target's timeline forward in time so that some events that you previously would have considered to be a part of your target's past are now, from your new perspective, in your target's future. Move away from your target fast enough (>c/10) and a sudden decision to deploy your 10c apparatus will send you to what you previously regarded as the past of your target. To shake your own hand, you just need to turn around and do the same procedure back to your starting point.
> Also, will Bitcoin outperform the US stock market over the next five years, at 51%. I started out thinking - of course it's 50-50! By the efficient market hypothesis, if any asset was obviously going to do better than another, people would change the price until it wasn't. But on second thought that's wrong - stocks have a higher than 50% chance of beating treasuries over the same period because of a risk premium. Maybe there's no intuitive way to think about this, you have to have opinions on the underlying fundamentals, and it's only 51% by coincidence?
Yeah, this is wrong, and I was also tripped up by this a while ago. The efficient market hypothesis only tells you that E[X] = X. But, e.g., if Bitcoin could only either go up 3x or crash to 0, then the efficient market hypothesis would tell you that p * 3 * Initial capital + (1-p)*0 = Initial capital => p = 1/3 = 33.3% of going up and 66.6% of going down.
It looks like you're still collecting play-money prediction markets but haven't mentioned The Foresight Exchange, one of the oldest. (I tried to post this last week.)
Perot did not win 5+% of the vote as a third-party candidate in 1992 and 1996. He won 8% of the vote as a third-party candidate in 1996, after having won 18% of the popular vote as an *independent* candidate in 1992.
It's an important distinction, because running for President as an independent candidate implies that you're trying to pull the Gabriel-over-the-White-House trick, riding into Washington on a white horse and sweeping away the corruption with the awesome power of the presidency. If you envision *not* being a dictator and actually sharing power with other people, then the first step is almost certainly going to be finding a bunch of other people (particularly wannabe-Congressman type people) with similar goals and getting organized.
Metaculus is scoring the prediction as positive if either a third-party or independent candidate reaches 5%, so at least they're aware of the distinction. But they don't seem to think it is an important distinction, and I think it's important to push back against that.
I think to most people, running as an independent is just a signal that you don't like political factions. You can refuse to identify with a faction while still respecting separation of powers and checks and balances, working with existing factions, and just generally not being a dictator.
I don't really think it's a meaningful distinction for the purposes of the prediction.
My criticism was going to be that only looking at the last eight (at the time of prediction) elections was cherry picking.
BUT it turns out the a third party or independent candidate has received >5% of the popular vote in 11/58 (19%) presidential elections (again, at time of prediction), which isn't *that* far off from 2/8 (25%).
What exactly is the distinction? If I win the presidency after running as an independent, is there any difference in how I am allowed to exercise my new Presidential powers compared to if I had won as a third party?
To a first approximation, the president has zero power except insofar as his allies in congress pass budgets and legislation to support his plans, and his staff in the White House competently and loyally executes those plans. Assembling a legislative coalition and a capable executive staff is close enough to establishing (or co-opting) a political party as makes no difference. If instead you think you're going to walk into the White House as an outsider, appoint a few of your outsider friends to the cabinet, and actually accomplish anything, then no, that's not going to happen.
None of that is a distinction. Is there a law preventing independents from forming coalitions? Will everyone in Washington refuse to talk to you if you declared as third party? What can an independent president do that a third party can't?
There's no law preventing someone from winning a (hot) war without an army, or manufacturing a million automobiles without a corporation. In theory, you could own a car factory as your personal property, hire thousands of people to work for you personally and tell them all to go run your personal factory, etc. But the organization that can build a million cars is a corporation whether you call it that or not. The force that can win a war is an army whether you call it that or not. And the coalition that can govern a nation is a party whether you call it that or not.
There *are* e.g, ballot access laws that make it significantly harder for you to do this if you don't call your coalition a "party" and fill out the paperwork as such.
And, none of the people who have run for the presidency as non-partisan independents in the past century or so, have made any real attempt to build such an organization. It's not something you can do in the two months between election and inauguration. Perot tried to do it after his 1992 presidential bid, but appears to have lost interest and phoned it in come 1996. "Independent" presidential candidates aren't people who have some perverse aversion to using the 'P' word to describe their thing that is a party in all but name. They're people who expect to play the Gabriel-over-the-White-House game.
"And the coalition that can govern a nation is a party whether you call it that or not."
Counterpoint: In Canada we have more than two parties that routinely win seats, meaning we often end up with a minority government that has to make deals with the other parties to get anything done. Sometimes you get a coalition that lasts a few years of party A and B working together to do all the things they agree on, but sometimes you get party A teaming up with B to pass environmental regulations one week, then teaming up with C to bust unions the next, over B's protests. Sometimes the MPs even vote against party lines which really muddles things.
If all governing coalitions are parties, the shifting nature of our coalitions either has new parties routinely popping in and out of existence, or people are leaving and rejoining one superparty on a weekly basis depending on what's up for a vote. This seems like an odd and not especially useful definition of "party".
But why is it *Party* A & B working together to do these things, rather than Party A + Party B + that guy who used to be part of party B but realized he could negotiate for things more specifically tailored to his constituent's needs if he went independent? And then next year just a hundred seventy independent legislators each seeking their personal optimum outcome? Legislators give up the ability to seek some things valuable to themselves and their constituents when they join a party. Why would they do that, if "party" is a silly meaningless null-word?
"Party", means in this context "organization actually capable of entering into the sort of negotiations you describe". Hundreds or even dozens of independent legislators, really can't do that. That's why they form parties, and that's why "party" is a useful word.
I think there's some value in distinguishing between consensus among those who are most informed ("expert consensus") vs among people in general. Maybe it's a bit redundant but it is not without information content.
A probability distribution over case counts, rather than a fixed-point, would have given most of the due credit here. The only downside is that a well-calibrated probability distribution requires much more thought.
Having lived in China, I regard their numbers as suspicious. Known to me is that if you're in the West and you die of an MI while COVID-positive you get scored a "COVID Death" whereas in China you're a "Heart disease Death."
Moreover in 2020 China didn't lock down Wuhan until After the annual Great Migration for Spring Festival had started. Many of my fellow ExPats had returned to their home countries, with varying local rigor of quarantine. ExPats or none, an annual migration of a billion people to and from the largest cities on Earth is not conducive to a quarantine.
My opinion based on my limited knowledge is that China's death and disease toll, were it counted the way we're counting, would be similar to India's given the similar living conditions of their poor and minority populations.
But of course, as a Westerner, I Just Don't Understand China—so say my coworkers—so it's entirely possible that regarding this ONE thing the Chinese government is NOT withholding embarrassing data that would cause them to "lose face" internationally.
But then you'd want to score a lie where it's clear what the truth is, so you can see the deviation. And you'd want to present it as a prediction on that, rather than on COVID.
If you have all three measurements, then sure, you want to test your prediction on all three. But when you know that all you have is the combined noisy measurement, testing your prediction on that is better than testing nothing at all, or testing your prediction of the thing that you claim without measurement.
In my experience, going only places where Westerners are reasonably allowed to go is that China's poor whom I observed were living in unthinkable squalor. You won't see that written about in Xinhua, but like Shanghai Pride, also not reported by Xinhua, nonetheless exists. Colleagues who have been assigned to both countries tell me that China probably wins on sanitation and violent crime.
Yeah; China has many poor regions; it is about as rich a country as Thailand. But both Thailand and China have been foresighted about quashing COVID outbreaks. The Hebei outbreak is over, and the Thai outbreak is becoming increasingly confined to Samut Sakhon.
"would be similar to India's given the similar living conditions of their poor and minority populations"
Impossible. China has been far more eagle-eyed about stamping out infections than India. Seroprevalence surveys have been consistent with the official death toll in Wuhan.
In grad school I took a course on probabilistic risk where our tests were graded using log odds scoring. Instead of picking A, B, C or D, you wrote in your probabilistic confidence in each choice. For example, if you’re somewhat confident in A, write in 60%, and then distribute the remaining 40% of weight across the other three choices.
Because of the log scoring rule, being very confident and wrong would result in a massive penalty. There was one star student in the class who blew the curve for all of us by putting down 90%+ confidence in his correct answers because he knew what he was doing and, furthermore, knew that he knew what he was doing.
This method was really hard and I got a pretty mediocre grade in the course, but I’ve always thought it was a good idea. You really learn about your own understanding of the material this way. It’s much easier to notice gaps in your own understanding when, instead of being given partial credit, you’re actively penalized for undue confidence.
A 3rd party POTUS candidate reaching 5% is not a particularly high bar nor, during my lifetime at least, a highly unusual event. It happened in 1996, 1992, 1980, and 1968. Three of the last 11 presidential elections and four of the last 14. So even before considering anything specific about 2024, I'd take the "yes" side on 15% odds.
On the third-party candidate: I think that the naive assumption of "Last 3 of 8" making the 15% bet worthwhile is a bit simplistic of an analysis. The first thing to consider is that, for such a small sample size, if it's possible to get more granularity, one should. And we certainly can. The political environments of those 3 third-party successes, so to speak, were not nearly as polarized as things currently are. Unless you think there will be a de-stressing of tensions (Possible) or a fracturing of parties (Very unlikely, given the two-party system's strength) I think it's closer to 1%. If there's an easing of tensions, call it 20%, very generously. Take this as me predicting against this.
Incidentally, it feels to me that there is some serious ambiguity on the "second wave in fall" question. Some would say the second wave was in the summer, and the later one was a third wave. Some would say that third wave was really in the winter, because it peaked in December and January most places. Some would say that although the summer wave seemed distinct from the first wave, it was really just a delayed first wave that hit Florida, Texas, and Southern California as soon as the initial restrictions were eased. Some would say that in the northern plains region, the wave that started building in August/September was actually just their first wave. Though I think that New Mexico and El Paso were clearly in a second local wave by October/November.
Here is an interesting paradoxical problem with this type of prediction self-scoring: suppose you are especially adept at thinking up prediction questions that are "hard" precisely because they successfully "cleave (potential future) reality in half". For example, your question 9 asks if China will have 100k deaths. In reality it had 96k deaths, so the threshold in this question is a pretty good guess!
If you were insightful enough to keenly set the thresholds for all your questions in such a way, you would frustratingly find that there is no way to earn a positive score at your own game (or even have much fun playing it), since an optimal Bayesian agent should predict 50-50 for all questions.
Re: "will Bitcoin outperform the US stock market over the next five years"
Consider my new cryptocurrency, Gamblecoin. On Jan 1, 2022 I will roll a die, and if that die comes up 6 it activates a smart contract that will buy anyone's Gamblecoins for $1 USD each, otherwise nothing happens. The odds of Gamblecoin beating the market are one in six, and this is not an efficient market violation because if it does beat the market, it beats it by a lot.
On a long enough timescale, Bitcoin should either become some kind of enormous global currency worth way more than it currently is, or become basically worthless as we find something else to serve the "anonymous secure global currency" role (doesn't have to be crypto, maybe the banks get their act together and come up with an efficient way to move USD) and Bitcoin no longer has a reason to go to the moon. The odds of Bitcoin becoming an enormous global currency are probably not 50/50, so there is some time period over which prediction markets shouldn't say it's 50/50 to beat the market.
That being said, there's a decent chance that Bitcoin is still in the speculative bubble phase and we won't know whether it becomes the massive global currency until far in the future. "Does Bitcoin beat the stock market over the next five days" should be 50/50 on the efficient market grounds you mention because the next five days are (very probably) just going to be speculative trading rather than a resolution to the question of whether Gamblecoin has value or not.
Vaguely on topic: somewhere in a thread on SSC or a related site, I predicted 1 million or more excess deaths in the US during covid, explicitly including deaths due to untreated non-covid conditions, despair due to unemployment, etc.
A month or two after that, US covid deaths were still below 1/4 million, and the rate was dropping, and I felt like an idiot - but I'd made the prediction, as my first attempt at measuring my prediction results.
When I checked today, the number of explicitly covid deaths reported in the US is over 1/2 million (501,117), and my confidence in my prediction is much higher.
Excess deaths - i.e. number of deaths in each 12 month period compared to the last 12 months without covid - aren't tracked as prominently, or reported as swiftly, so I can't yet check how close we've come so far.
I'm reposting mostly to continue to hold myself accountable.
I see a number of people modeling bitcoin outcomes as binary. That may be reasonable, but just about every other financial asset is very well modeled as a random walk with drift (expected return). If we do that, stocks have about 15% annual volatility over long horizons, and bitcoin has... I actually haven't measured it but it's probably 5-10x that. Expected returns on stocks is probably in the ballpark of 4%, looking at international markets (US has done better, but using its returns to predict its future returns is probably worse than using the average international estimate). Over a 5 year period, the volatility of bitcoin is so much higher than stocks that the average return on stocks barely matters. So the random variable (bitcoin 5 year returns - stock 5 year returns) has a negative mean (if we assume bitcoin doesn't have a positive risk premium), but its volatility is so high relative to that mean that the probability of it being greater than 0 is pretty close to 50%.
Careful: if you assume a lognormal distribution of returns (which isn't empirically accurate, but it's usually the default approximation), higher volatility doesn't always translate to a higher chance to cross above a barrier. As volatility gets high, most probability mass concentrates around 0 in a lognormal distribution.
That’s a good point. I was being lazy about normal vs lognormal, and at these high volatilities it matters a lot. So I take back my comment - it actually may be reasonable given this level of volatility to say that it’s either zero with some high probability (or close enough to it that you clearly underperform equities) or much higher than it is now, and the probabilities around those two possibilities are not necessarily 50/50.
>16. Will there be a general consensus that summer made coronavirus significantly less dangerous?
Maybe this is true, but I have my doubts. I found this posted on Good Judgment Open:
"Given the lack of immunity to SARS-CoV-2 across the world, if there is an effect of temperature and humidity on transmission, it may not be as apparent as with other respiratory viruses for which there is at least some preexisting partial immunity. It is useful to note that pandemic influenza strains have not exhibited the typical seasonal pattern of endemic/epidemic strains. There have been 10 influenza pandemics in the past 250-plus years—two started in the northern hemisphere winter, three in the spring, two in the summer, and three in the fall. All had a peak second wave approximately 6 months after emergence of the virus in the human population, regardless of when the initial introduction occurred."
I will add that the Winter Is Coming! narrative probably seemed like convenient messaging to some of the public experts and encouraged the perception that there was something special about winter that made the 2nd wave so severe. It fed into public intuitions about winter being worse for the cold and flu season.
But if we go with the above notion that the 2nd wave was pretty much bound to happen regardless of the season, it seems more a matter of luck that the 2nd wave did coincide with winter. Of course, that doesn't answer the question of whether the 2nd wave coming in winter did make it significantly worse than it would have been otherwise.
My intuition is that 2nd waves in pandemics are likely to be worse. Unlike the 1st wave, there is no "patient zero" in a given city who orbits within a few social circles. As the 1st wave subsides and R0 waffles from >1 to <1 to ~ 1 in a given local, the disease spreads gradually to people in all walks of life, orbiting all social circles, in all communities, so as the 2nd wave starts due to say a general relaxation in social behavior, we now have a thousand "patient zeros" in a thousand locations, a thousand sparks able to ignite the kindling in a thousand more locations, bringing the starter wood to a hotter temperature and ultimately burning the forest with more ferocity than the fire could six months prior.
"Most papers I read now agree that coronavirus is a seasonal disease and that it's not a coincidence that cases went down in summer and up in winter. "
Because I also tried to hobby-research this question just a few days ago, and came to exactly the opposite conclusion. The papers that convinced me most found a non-zero, but pretty minor effect, like a ~10% increase in the R-value if temperatures rise by 20°, which is roughly the difference between European winter and summer. I also looked anecdotally into the numbers, and most European countries had a quite strong and consistent exponential growth during all of summer.
It seems that you can find a correlation between "winter" and "high numbers" in a lot of countries. As Jack Wilson points out, this is probably coincidence and tricks intuition of many people. Actually, this connection doesn't make any sense. It should be between "winter" and "high growth rate", and *that* connection does not seem to exist. It does exist for influenca, which starts growing around September/October and doesn't reach high numbers before January/February, but corona doesn't seem to follow this pattern.
I might add that I find it possible that covid-19 does become seasonal in the future. Once immunity in the population keeps it just below the critical value, then a 10% increase may become the push that brings in into supercritical regime. But I think we are still far away from this point.
These markets seem to be modeled on, well, markets, but the function they aim to fulfill seems to be closer to that of a consulting firm. Betting/investment markets, at least in the US, come with a prohibitive degree of red tape.
So, bizarre question: would it be possible to create a prediction "market" that functioned like "the Uber of consulting firms"? Clients contract with the firm for a given amount with a set of things they want predictions on. "Independent contractors" can buy a percentage in the contract and submit their prediction(s) for the associated questions. Payout is a function of accuracy and percentage of buy-in.
1) Could that work?
2) Would it be legally less difficult than prediction markets as currently constituted?
Re Third party 5% share: Seems like the easiest path would be a Trump Patriot Party-esque splinter. This'd require Trump to rerun (or find a charismatic substitute) and for the GOP to resist hard enough to create an election-losing vote split over it, which seems iffy, but certainly raises the chance beyond the background chance in the previous elections.
Last year, I built a site that was based on predicting the future. (Don't worry, it's dead, this is not an advert)
The idea was to make it a more general, lower-entry prediction market. Instead of people betting money, they would gain or lose reputation. The site itself would function like a geopolitics focused subreddit, except that people with good reputation have much stronger upvotes. Therefore, the smart people would rise to the top much faster, and if a strong expert made a comment a few hours too late, it would not be buried. Second, any unproductive groupthink or echo chamber effects would be broken down quite fast - the people that succumbed to this sort of stuff would lose reputation, and the people that didn't would rise up, relatively speaking.
And of course, the reputation of every poster would be visible next to their comment for bragging rights.
In the end, I couldn't figure out how to reach a critical mass of users and took it down; here is a mock-up I used that shows how it worked, if anyone is interested.
Gladly. Please check the following two publications demonstrating that planetary positions correlate with phenomena with high r2 values written by Renay Oshop my project partner.
"Astronomy of the day is more effective than seasonal decomposition in modeling and predicting the rates of Amazon review misspellings," International Journal of Scientific Research and Management, Feb 2019, Vol. 7, 2, DOI: 10.18535/ijsrm/v7i2.aa01, https://ijsrm.in/index.php/ijsrm/article/view/2040/1719
On third party vote share: Gary Johnson got 3.27% of the popular vote in 2016. I think if you re-ran the tape a few times he might have crossed 5%. Indeed, FiveThirtyEight's final forecast (https://projects.fivethirtyeight.com/2016-election-forecast/) had him at 5.0%.
Talking to some people on Twitter suddenly reminds me how, as a Gen-X er, I'm still bitter about how much the government in the form of Surgeon General Dr. Koop scare mongered straight people about AIDS. They wanted you to believe circa 1990 that all oral sex was going to give you AIDS if you didn't wear a dental dam.
I'm not saying that mattered for most people, but for many of us on the margin I think it was a big deal. Sex is awkward when you are young and the fear of AIDS made it much more awkward. Not much is written about that these days, but when I think back it was a huge deal at the time.
Agree. Aids was so scary back then with magazine PSAs showing an attractive/available woman an then telling you on the next page that she actually has aids and you could die if you had sex with her, - "You're not only sleeping with her, your sleeping with everyone she's ever slept with any everyone they've slept with too." Also a mirror page showing you a "typical aids victim". While I was losing my virginity, I was thinking about Aids infection - "I hope to god this condom doesn't break or any fluids leak round the side, I know she's only had one boyfriend before but how many people has he slept with...??? I don't want to contract HIV and die of aids. Fuuuuuuu". Not conducive to a fun relaxed first-time shag.
How do we know if we over- or under-reacted? It seems like a lot of people are locked in to defending their views on what the right level of reaction was.
By the way, here's a question for helping come up with an objective assessment of how bad covid was. The Spanish flu of 1918 hit men between 20 and 40 the hardest, thus leaving a lot of widows and orphans. With 500,000 Americans now having died from covid, how many orphans (under 18 years old) did this toll create?
Just wanna say it's so strange that neither you nor your friends got covid-19. I am russian and it highlights again, how different is our risk tolerance, comparing to the rest of the world. I remember one link you posted earlier, where some guys try to calculate the risk of some some interactions, going from the assumption that the reasonable risk of getting covid per year should be 1%. For me it's ridiculous. I would say 70% chance of getting covid per year is ok, maybe 40% if you are extra-cautious.
I got covid (asymptomatically), my wife got covid, my sister and her husband got covid, my wife's cousin and all her family got covid, my best friend got covid (asymptomatically) and his mother got covids, some other friends also got covid. (No deaths, one hospitalization).
And Russia has so few restrictions. After spending five months last year in Chile in a lockdown I enjoy the freedom to leave my house tremendously.
All response are a matter of conjecture. Posit if you will. Nobody I have seen in the public format is a prophet; it's a good thing to give and take in an opinion which can be refined to it's essence without another opinion.
I made predictions last year, inspired by Scott. I got 7/8 95% ones, 15/16 90% ones, 10/13 80% ones, 6/12 70% ones, 8/12 60% ones, and 2/12 50% ones. It seems like my largest error was in the 50% ones, strangely enough, I think due to my tendency to throw a lot of low-probability things with affirmative phrasing ("this WILL happen") there. As a fun fact, the 95% one that I got wrong was "I will still regularly be reading SSC at the end of the year: 95%"; this was incorrect for reasons entirely different than I imagined.
In retrospect, it almost seems like writing questions - in particular, anticipating all possible outcomes and then phrasing questions so there will be unambiguous resolutions given any possible outcome - is harder than assigning probabilities once you've already written the question.
I argue that 15% is on the lower end. The relevant base rate is near 20%, and four years is a long time in politics. I'd say it is too long to make strong predictions about specifics of he next presidential elections to deviate from the base rate more than 5 points.
Partisan tension may be high, but I'd argue that heightened tensions and general weirdness of recent times contribute towards general uncertainty about electoral results. In addition to established third parties, there is a non-zero chance for large scale party realignment or splits, and I can see substantial part of uncertainty splitting in that direction. Was there much reason in 1908 to suspect the great upset of 1912? There had been only one 5+% third party election in previous ~30 years, there was little indication that TR would run *against* Taft or that Debs would double his 1908/1904 level turnout. As a case of upset of smaller magnitude, Perot launched his 1992 campaign only after the primary season had started.
I would pick the general base rate as the best estimate that incorporates how often this kind of things may happen in US politics. And anyway, it still gives 80% chance for all 3rd candidates getting <5% of votes, which is the usual trend. 80% is often.
China's official case count on 12/31/20 was 95,963, so false."
Actually, depends. China publishes two different offical numbers. What you took is the number of symptomatic cases (which China, other than all other countries, enters into the official statistics websites). China also publishes the number of asymptomatic cases each day (positive test but no symptoms), but doesn't make them available in a nice and accessible format. If you count those, too, you are way above 100,000.
Long thoughtful article by Siddhartha Mukherjee. "While the virus has ravaged rich nations, reported death rates in poorer ones remain relatively low. What probing this epidemiological mystery can tell us about global health."
"I will never bet against Zvi on anything. Last year he bet against me on what restaurant I would have dinner at, without knowing anything about my situation or food preferences, and won anyway."
> will a third-party candidate win 5%+ of the popular vote in 2020?
I got very confused before clicking the link: this should be 2024
Odds on that 3rd party candidate being Trump? (core Republicans 'get their party back', Trump runs 3rd party)
Here's my dumb prediction: Neither Trump, any member of his family, not any protege closely and personally linked to Trump will be a serious candidate in 2024. 75%
Because each day in coronavirus lasts 100 years, this may not seem true, but Trump was POTUS 34 days ago. His relevancy will fade a lot in the next several years, and people who currently claim that he's the best ever will quietly shift their opinions about him.
To me, "not a serious candidate" means that the person won't be on the GOP ticket, and if they run 3rd party they do not break 4%.
If an immediate family member of Trump runs, they'll break 4% without even campaigning. The Cult 45 crowd will get them that far on unquestioning loyalty alone. Even four years from now.
Maybe! But I think people underestimate the strength of the two-party system, and overestimate people's current protestations of loyalty.
I mean, my prediction is 75%, not 99%. That's because of two things:
1. I do think that Trump has an unusual amount of personal cred with some elements of the GOP.
2. Also, the media on all sides is addicted to talking about Trump and extending his relevance.
Still, I overall think that a huge percentage of the people who right now are like, "OMG, TRUMP" are gonna be, in 2023, "Yeah, I was always kind of dubious about that guy."
I tend to agree with you. The political landscape 4 years from now is very hard to predict. Imagine calling that Trump would win the presidency in 2012, or that Obama would win the presidency in 2004. That would be foresight at a deific level.
In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies. This is very different, but much more attractive, than Trump's previous disconnected constellation of positive self-image affirming tendencies.
It's also worth pointing out that the potential for wit of a given candidate is heavily dependent on who is advising them. Observe Trump's 2016 campaign with his 2020 one. His advisers back then were of much higher quality, and I think that's in part because of the transient power he appeared to offer.
What I mean by this is that, keep an eye on political players that drift just below the surface of figure heads like presidential candidates.
>In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies. This is very different, but much more attractive, than Trump's previous disconnected constellation of positive self-image affirming tendencies.
I'm not sure it is more attractive to a large proportion of the base who supported Trump, because I don't think a cohesive set of policies was a criterion they paid much attention to in the first place. If anything, it might have ended up being a burden to his popularity, if it gave too clear a way to judge how much of his platform he actually managed to accomplish.
I think there has been a tremendous overreaction to Trump's win and subsequent predictions of what the next surprise success story is going to be. Trump barely won a clown car nomination process with a minority of support and the open disdain of much of the party and conservative media apparatus, which then immediately fell in line and started telling themselves stories about why he appealed to them when the only reason he appealed to them was he won and had an R next to his name. The sheer level of partisanship in the country is going to lead to wild swings in party profiles as voters and pundits try to convince themselves they have actual reasons to support whoever gets lucky at the right time in crowded and long nomination battles. It probably won't lead to much in the way of policy or governance differences just due to the intransigence of Congress and the judiciary, but it'll make the characters of each party seem like they've shifted a whole lot between 4-8 year periods.
Plus, for the people who legitimately loved Trump from the beginning and didn't just change their minds once he won the nomination, too much of his appeal is tied up in his unique personality and celebrity. He's a billionaire who married a model and fucks porn stars on the side. There's a "man, I wish I could live my life like that" quality that boring nerds like Tom Cotton and Josh Hawley can't replicate even if they're more ideologically coherent.
"In general I think it's much more likely that a younger and wittier candidate will emerge that fuses elements of Trump's populism with a subset of cohesive policies."
This is as likely as Forum for Democracy getting anywhere. Which is to say, it isn't likely. Trump didn't win because he was a "populist", whatever that means, anyway, he won because he was a rich businessman, which caused the low IQ to think he was good on the economy.
The linked Metaculus prediction was 15%. Your 75% prediction that Trump or a family member doesn't break 4% is compatible with a 15% prediction that they or someone else breaks 5% as a third party candidate.
I was basically trying to differentiate between "unexceptional third party candidate support" and "exceptional (but still doomed) third party candidate support." Without looking carefully, my impression is that in a typical year, there are third party candidates who get about 3% of the popular vote, so more than 4% is unusually high.
I agree with this, but for different reasons. I think hard as this may be to believe that it's Trump himself who won't be interested in running. Indeed, I think he wasn't interested in being President by the end of his term. His entire trajectory from about late October to the end was *not* that of someone who is (1) fairly intelligent and (2) genuinely interested in maintaining political power. I don't think he's stupid.
So why did he *say* he wanted to keep power? We can't fully discount his personal enjoyment of kicking up a fuss, because he does like that. But my suspicion is it was all about the money. He is under very serious financial strain, and by provoking a crisis and selling his true believers on the urgent, urgent necessity to go all out to respond he raked in something like $250 million between election day and the Inauguration. That will go a very long way to easing his immediate personal financial difficulties.
So while I don't think we should discount the possibility that he appears in various places to fling grit in the gears, because he likes that, I think he will find some reason (or engineer it) such that he doesn't personally run ever again. I think it's kind of been there done that, for Trump personally.
Thanks, fixed.
Scott, didn't you link to papers showing India had almost as many people infected by now as the entire U.S. population? India obviously has had more COVID deaths than the U.S.; official deaths seem underestimated (conservatively) by 10x. So this must be wrong:
"I haven't seen these numbers, but I haven't seen anyone make a plausible case some other country is off by a factor of two. This was mostly in there to cover China (or someone) lying about their death count in an obvious way, which didn't happen. True."
Indonesia might also have more COVID deaths than the U.S.; its official figures are ridiculously underestimated.
I don't know what papers you mean. I thought I saw India has 6% seroprevalence, which certainly means fewer infected than the US population.
That was in August. Months pass by. Let's go by that paper's estimate that true infections were 26-32x official cases (and there's no reason to think that ratio has changed). Official cases on Dec. 31 were 10,266,674. This implies between 267 million and 328 million true infections and between 1.3 million and 3.3 million true deaths. This is surely much higher than the American total. This is consistent with a later seroprevalence survey between December 17 and January 8 which reported 21.4% of adults surveyed had been infected with coronavirus, which implies total infections (assuming the same % children were infected, though it was reported higher) of 292 million infected, which would imply between and 1.5 and 2.9 million deaths by December 31. That's a lot more than in America at the time.
https://science.thewire.in/health/third-national-seroprevalence-survey-icmr-covid-19-rural-prevalence-test-positivity/
https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(20)30544-1/fulltext
I think you're right, thanks. I'll mention this next Open Thread.
Where do you get your IFR from? Why do you think the real death rate is 10x the reported, let alone "conservatively"? Your first link says that based on age distribution, you should only expect 3-4x, so a death toll equal to or less than the American. There's a lot of uncertainty, so it could easily be over, but your own source says it's under.
"Where do you get your IFR from?"
Peru and Mexico, which have a median age barely above India.
Even ignoring risk premiums the efficient market hypotheses doesn't predict a 50% chance that one asset will outperform another. It predicts that any two assets will have the same *expected* returns. But if one of the two assets has an asymmetric risk profile (for example a large probability of doing worse balanced out by a tiny probability of doing *much* better), the chance of doing better might not be 50%.
Yeah, I was going to point this out. I wonder which direction the asymmetry goes in the case of Bitcoin. At this point I wouldn't be shocked if it rose meteorically or crashed horribly (or both).
Re question 11: Hdroxychloroquine is not a particularly effective treatment for covid. It does, however, seem to be remarkably effective as a prophylaxis. This is likely how India beat covid against all odds. India accounts for three-quarters of the malaria in Asia and large swaths of its population take hydroxychloroquine regularly because it is less expensive than more recently developed antimalarials.
The evidence I saw for HCQ as a prophylactic seemed weak and post hoc, and hasn't India done worse on COVID than any of its neighboring countries?
You are the physician, Scott. I'm just an engineer. But my understanding is that India has beaten covid when we expected it to be devastated. Some say it is masks, but they don't wear masks. Some say it is genetic immunity with no evidence. Some say it's just because most of India has no sanitation and therefore robust response to infection. No one credits the widespread use of HCQ there. Why? Only because Trump recommended it. I am honored that you actually read my comment.
I would also be shocked if any significant number of Indians used hydroxychloroquine regularly. Just because you live in a malarial region doesn't mean you take lifetime malaria prophylaxis pills.
I won't quibble, but I did read a statistic to that effect, although perhaps not on a reliable source. But the effect is real. So, maybe its due to the masks they don't wear.
You have long been my hero, Scott, and I don't want to argue with you about things of no consequence.
"So, maybe its due to the masks they don't wear."
They do wear masks.
It was my experience, visiting rural India recently, that most of the people there cannot afford masks. Literally. It looked to me as though maybe 1% of the people around Bangalore were wearing masks. And only 20% had eaten, that day.
Malaria is not widespread anymore in India. I visit often. The travel clinics in the U.S now say there's no need to take HCQ once a week prophylactically, unless going to specific parts of specific Indian cities. India might have beaten malaria in many cities, using DDT.
And mask wearing has been fantastic in big cities in India.
Covid numbers are edging back up slowly there again :( now.
"But my understanding is that India has beaten covid when we expected it to be devastated."
The Indian COVID pandemic was basically the March-May NYC outbreak stretched out over four times as long a time. Official Indian statistics dramatically understate cases and deaths, and I expect that to be the case in Bangladesh and Pakistan as well.
"but they don't wear masks"
Yeah; they do, India has one of the highest mask-wearing rates in the world: https://d25d2506sfb94s.cloudfront.net/cumulus_uploads/inlineimage/2020-04-22/face%20mask%20usage%20by%20market.png
I've been using India as a baseline for judging the US.
Basically, it has like 4x the population, densely packed cities, widespread poverty, lots of international travel, but not a particularly powerful central government (unlike China). We should naively expect India to have a worse time than America. Since India has 150k deaths compared to America's 500k, we can estimate that at least several hundred thousand US deaths are due to America-specific failures (as opposed to "too much rural population" or "price of democracy" or something).
I know this is a massive oversimplification, but it sounds like you have a better understanding of the Indian situation. How much might India's death count be off by? Is there an obvious reason their death count is lower that can't be attributed to a failure of US government/civil society?
"hasn't India done worse on COVID than any of its neighboring countries?"
Nobody really knows what India's neighboring countries' true COVID cases/deaths are.
People may be underestimating how good hydroxy is, but the question was if there would be a "consensus" that it's good, and there definitely isn't.
People in India get the BCG vaccine for TB. If there is a biological rather than social explanations for less COVID deaths in India it is probably permanent priming of the immune system by BCG. BCG results in a lot of interferon production which might enhance both innate and specific anti-viral responses.
I don't think it's necessary to conclude that you and Bucky were "worse" than guessing 50-50 on everything. For example, such guesses would imply a distribution with 0% of probability mass between 100k and 3M US COVID deaths. The predictions look pretty reasonable to me; I know that a lot of people (myself included) were underconfident about the possibility of a 2nd wave.
I also think that given the actual official case number from China, that question should probably be scored higher the closer it was to 50%.
Also, isn’t it weird if the scoring rule doesn’t penalize you for answering 50% on three questions about the same number at three different levels, separated by orders of magnitude?
Is it possible for the real probability of X > Y to be 50% at the same time for Y = 100k and Y = 1M? I guess we could be in a world where future X is either 1E9 or 1E3, with 50% probability, in which case the two answers are both perfectly fine. But it still feels as if there should be a penalty there in the real world. What am I missing?
Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths. There's almost certainly a lot of undercounting going on (though it may not be deliberate lying). The UK, with their unusually loose standard for a death within 28 days of a positive test being a Covid death, is one of the only nations where COVID deaths and excess mortality are similar - in fact the UK has slightly more COVID deaths than excess mortality.
That still isn't enough to knock the US off the top afaiw though. But of course, China's figures are worth as much as a bet on Trump being president, but I suppose it doesn't qualify for being obvious and no one will ever know the true death toll.
> Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths.
This assumes that the only difference is covid. We know that isn't true; the societal reaction had wide-ranging effects.
> Pretty much every nation has lied about their death counts in an obvious way, since almost every nation shows much higher excess mortality than official covid deaths.
Not necessarily. Some of that may be deaths caused by the reaction to covid; I've at least read articles about suicides being significantly up, and people being reluctant to visit the hospital if something might or might not be wrong (which a lot of people were when the panic was at its worst - everyone knew the hospitals had covid patients) is going to give you some missed heart attacks and strokes. On the other hand, traffic deaths should probably be down - there's much less driving, even if what there is is sometimes more reckless. Last I heard we had anecdotal accounts but not a lot of actual data on the non-covid breakdown of the excess deaths, but that was some time ago and I would love to be proven wrong.
This might be true for some countries, I guess. But at least for the Czech Republic (where I’m from), our terrible total excess deaths are totally correlated (0.94) with our terrible covid infections, so no extra deaths to be seen. (Yet! unfortunately, reduced prevention will probably result in excess deaths in future.) For a chart, see e.g. https://boundedlyrational.substack.com/p/nadmrt-a-kauzalita
Surprisingly, suicides have significantly decreased since the start of the pandemic, at least in Canada and the US.
https://twitter.com/tylerblack32/status/1357765466601775106
https://docs.google.com/spreadsheets/d/1xt4C67x-gGwBl6eXU-ZNrZq9oQvwBdwjVHAzR5H7jtI/edit#gid=466591535
"That still isn't enough to knock the US off the top afaiw though."
It is. India has almost certainly had more COVID deaths than the U.S., and there is a significant chance (30%) Indonesia has had more. Russia is coming pretty close to surpassing the U.S., as well.
huh, consider me informed
India has almost certainly had more COVID deaths than the U.S
This is almost certainly not true.
Yeah; it is true. Official cases in India have been fairly consistently about 1/30 of infections. Even assuming the IFR in India is half that in the U.S., it still has a lot more deaths than the U.S.
Not sure why that assumed IFR is supposed to be reasonable. I agree that total infection counts would be a lot higher than reported but it seems very likely that serious/critical cases in India were still a smaller fraction than in the US. Across south Asian countries, deaths per million population are 10-20 times smaller than the US or the UK. I would assume some undercounting, but while it's easy to miss infections, hiding that many deaths (inadvertently or on purpose) doesn't seem probable. Based on (admittedly spotty) reporting, I would assume an upper bound of ~(2 x reported) on actual deaths in India. Here is one example of excess death reporting: https://science.thewire.in/the-sciences/covid-19-mumbai-all-cause-mortality-data-ifr-bmc-seroprevalence-survey/. It's not representative of the whole country but Mumbai had one of the (if not the) worst outbreaks in the country, which disproportionately hit slum areas (so probably should have a higher fatality rate than average).
"Across south Asian countries, deaths per million population are 10-20 times smaller than the US or the UK."
There is no reason whatsoever to believe that's actually the case.
"Not sure why that assumed IFR is supposed to be reasonable."
There is absolutely no chance Peru's and Mexico's IFR was lower than .5%. Peru's and Mexico's median age is very close to India's. Assuming the seroprevalence surveys in India are even close to accurate (and there is every reason to believe they are), India has had at least twice as many COVID deaths than the United States, and likely more than 3x as many.
Also, Mumbai is one of India's richest cities. In Russia, where deaths are understated by a factor of 6.3, they are understated by only about a factor of 2 in St. Petersburg and Moscow. I am confident that the reporting divide between core cities and the other parts of the country is even worse in India than it is in Russia, and India's core cities make up a much smaller portion of the population than Russia's do: https://raw.githubusercontent.com/dkobak/excess-mortality/main/img/regions.png
I am not convinced on the utility of comparing IFRs. It's hard to trust infection counts even in the US, where there has been a lot more RT-PCR testing than any of the other countries you mention. IFRs are normalized by total infections but that denominator does not reflect actual infections, and it's hard to say how off it might be in different countries. How do you account for the variable testing and reporting issues across countries?
Excess mortality reporting, although patchy right now, is more useful for bounding the total deaths. Several Indian cities and states conducted death audits to account for excess deaths and it shows undercounting of "COVID deaths" but nothing supports even a factor of 5 undercounting.
You are making confident, numerical claims. Please bring evidence.
I did.
The only link from you in this subthread is about excess mortality in Russia. You've maid claims about India (such as that official cases are 1/30 of infections, and claims about the number of deaths based on the number of infections), Indonesia, Mexico, Peru.
I have a pretty strong prior that any cause of death that disproportionately impacts old people will be undercounted, not out of malice, but just because we don't generally investigate too hard on cause of death when someone dies in their 70s or 80s, so unless they had a diagnosis prior to dying, it won't get counted. This means nothing for international comparison because all countries would have the same issue, but I am consistently surprised that people think Covid deaths are overcounted.
IIRC there were several months where the US was running COVID tests on every dead body just in case.
I'm almost certain there has never been a consistent nationwide policy on that type of thing. It's up to county level medical examiners to deal with that.
"Will the Bay Area stay locked down until Election Day (11/3/2020)?" was worded in your original post as "Bay Area lockdown (eg restaurants closed) will be extended… until Election Day". Both wordings to me are unambiguous: they imply a single continuous lockdown. If they meant multiple lockdowns, the former would read "Will the Bay Area be locked down on Election Day" and the latter would read "Bay Area lockdowns will continue on and off".
Zvi also explicitly clarified on his post (https://thezvi.wordpress.com/2020/05/01/slatestarcodex-2020-predictions-buy-sell-hold/): "If anything, that seems high, assuming it means continuous lockdown until then rather than being locked down on election day." I think omitting that result from his score would be better than including it, because he was explicitly predicting a different thing from you.
I don't think restaurants ever opened for normal business before Election Day. I realize this was ambiguous, but I was trying to get at the idea of "when will the pandemic be over".
I think the issue of lumping the Bay Area is part of the problem (https://abc7news.com/timeline-of-coronavirus-us-covid-19-bay-area-sf/6047519/) Different counties had different experience of lockdown but the original orders remained in effect in most of the counties through today.
Incidentally, I've really wished that no one would use the word "lockdown". It seems to mean totally different things to different people. My friends in France were telling me a few weeks ago that restaurants and shops are closed, and you're not allowed to travel more than 100 km from your house, but the government is saying there's still a possibility that they will go back into lockdown. Meanwhile, in Texas, people think we're still in lockdown because you're required to wear a mask when you walk into a restaurant until you sit down.
I'm getting different totals for the scoring. Could you/someone else check to see what's going on? Just to make sure we're doing the same calculation: Let's call p the probability that a forecaster assigned to the correct answer, then LogOdds = log(p/(1-p)).
Some examples: scott's score on the first question was log(0.6/0.4) ~= 0.405. His score on the third question was log(0.9/0.1) ~= 2.197 (balanced out exactly by his score on the second question).
Doing this computation for all questions gives total scores for Scott, Zvi, and Bucky at 7.2034, 13.2863, and 6.5006 respectively, all better than chance!
Also, I recommend switching to Brier score (squared error) for things like these. It's a proper scoring rule, which means that you cannot game the system; in expectation you'll get the best possible score by reporting your true beliefs. In contrast, something like LogOdds can be gamed since increasing the extremity of your forecast just ups the ante, but the odds of the bet don't change [i.e. log(p/(1-p) = -log((1-p)/p)]. For instance, if you think something is 90% to occur, then by reporting your true beliefs you are expecting to get 1.7578 "LogOdds points" (0.9 * log(0.9/0.1) + 0.1*log(0.1/0.9) ). However, you could expect to score more points by reporting a more extreme belief, like 95%. In that case, you would expect to score 2.3556 points (0.9 * log(0.95/0.05) + 0.1*log(0.05/0.95)). Doing the same calculation with brier score for belief p and *reported* belief x has you minimizing the function p*(1-x)^2 + (1-p)*x^2 for values of x, which occurs only when p=x (note that lower Brier scores are better).
Here's some Matlab code to run these computations:
answers = [.6 .4 .8; .1 .1 .1; .1 .05 .05; .5 .3 .6; .9 .9 .95; .8 .9 .8; .7 .8 .6; .9 .95 .8; .7 .4 .7; .5 .4 .4; .2 .15 .6; .3 .2 .1; .6 .4 .3; .5 .3 .6; .2 .2 .1; .7 .7 .4; .3 .2 .2; .9 .8 .9; .2 .15 .3; .2 .1 .2]; %Scott Zvi Bucky
outcomes = [1 1 0 0 1 1 1 1 0 1 0 0 0 0 0 1 1 0 0 1]'; %Scott's recorded outcomes to the 20 questions
err = abs(answers-outcomes); %Compute error for all answers
accuracy = 1-err; %The probability assigned to the correct answer
brier = (1-acc).^2; %brier is squared error
LogOdds = log(accuracy ./ (1-accuracy) ); %Compute LogOdds for all responses based on accuracy
sum(LogOdds) %display total LogOdds for each forecaster
mean(accuracy) %same for average accuracy
mean(brier) %same for average brier
%% How might a well-calibrated forecaster improve their score?
p = 0.9; %well-calibrated belief
x = p; %stated probability. For now, say what you believe
(p*log(x/(1-x))) + ((1-p)*log((1-x)/x)) %compute expected LogOdds points
x = p+0.05; %say something a little more extreme
(p*log(x/(1-x))) + ((1-p)*log((1-x)/x)) %compute expected LogOdds points
%Compare with brier
x = p; %stated probability. For now, say what you believe
p*(1-x)^2 + (1-p)*x^2
x = p+0.05; %say something a little more extreme
p*(1-x)^2 + (1-p)*x^2
He used the logarithmic scoring rule for the calculations (calling it "log odds" in the post is an error). The score on a question is just ln(p) - ln(.5). This is a proper scoring rule.
When I do that calculation I get a total of -0.54 for Scott, 0.81 for Zvi, and -1.95 for Bucky, which is identical to what Scott reports for two of the three but slightly different for Zvi's score.
Yeah, when I use logarithmic scoring I get the same results as you do - so Scott's calculation for Zvi is slightly off and his description is also a little wrong. Thanks!
Thanks, fixed.
Forgive my ignorance for question 12, "Will I personally will get coronavirus": isn't a 30% probability of this being true equivalent to 70% probability of it being false... i.e. Scott correctly predicted he wouldn't get covid?
Questions marked red are resolved as "false", black ones as "true". Predictions are then scored based on whether they're high probabilities (if the question is black / true) or low probabilities (if the question is red / false).
Scott predicted he wouldn't get Covid with a probability of 70%. The logarithmic scoring rule then says that Scott does deserve credit for this prediction (which contributes a positive term to his overall score), but somewhat less than Zvi (80%) and Bucky (90%).
> I don't know how else to visually represent "mantic"
By a "mantle" (i.e., pallium). :-)
Looks like "mantis" does derive from the "correct" Greek word for this https://en.wiktionary.org/wiki/mantis#Etymology but "mantle" doesn't.
Something related to the Orcale at Delphi?
Next time perhaps include a competent astrologer who also uses ML. Maybe a female for stronger intuition. It would be v interesting to see the results.
"Competent astrologer" is an oxymoron
More than a professional gambler I take it?
Professional gamblers are quite well attested, see eg the final table at the WSOP often having the same set of people. I can't tell if this series of posts is a troll, though.
So are astrologers who work with AI to test their predictions. I assume you know a lot about how astrology works to be so dismissive. Otherwise we have the same issue as with UFOs...people closed minded (i talk about astrology so i must be a troll) which perpetuates prejudice and ultimately precludes brilliant people from studying it and possibly open up new avenues.
I will bet you $20 that you can't find 2 astrologers, with no personal or professional connections to each other, other than being professional astrologers, whose horoscopes will agree given the same anonymous person's star chart, and in the event you do find such a pair, I'll pay an additional $10 if they agree on whether Ophiuchus should be part of the zodiac or not. As for UFOs, there's a huge difference between "it's incredibly unlikely given the size of the universe that we are the only intelligent life in the universe" and https://www.amazon.com/Space-Raptor-Invasion-Chuck-Tingle-ebook/dp/B00S4B95RQ
This makes no sense. Literarily none. Do you know anything at all about astrology? "that you can't find 2 astrologers, with no personal or professional connections to each other, other than being professional astrologers, whose horoscopes will agree given the same anonymous person's star chart" I am a western astrologer. No one I know cares about Ophiuchus as constellations have no bearing on our work other than the origins of the historical names for signs. We work with the tropical calendar which is not anymore in sync with the current position of constellations.
I talked to a few professional gamblers, and the explanations of how they worked made perfect mathematical sense.
One of them was a cab driver in Las Vegas, who used to make decent money playing card games at Las Vegas casinos, until a bunch of (or all?) casinos banned him for card-counting. Apparently card-counting gives you enough of an advantage that no casinos want to deal with you if you're doing that. If he was less obvious about it, perhaps he would have still been earning an income by gambling.
Another was a poker player in an online casino. He explained that, first, he is a lot better than the average player in room, and better enough that it's worth playing (it being mostly the other people who absorb the average loss); and second, the online casino paid people like him to fill out tables that did not have enough players.
I'm going to regret asking this, but can you tell me more about astrologers using machine learning to check their predictions?
Jutka2 min ago
Gladly. Please check the following two publications demonstrating that planetary positions correlate with phenomena with high r2 values written by Renay Oshop my project partner.
"Astronomy of the day is more effective than seasonal decomposition in modeling and predicting the rates of Amazon review misspellings," International Journal of Scientific Research and Management, Feb 2019, Vol. 7, 2, DOI: 10.18535/ijsrm/v7i2.aa01, https://ijsrm.in/index.php/ijsrm/article/view/2040/1719
and
"Twitter Followers Biased to Astrological Charts of Celebrities," Journal of Scientific Exploration, Vol. 29, No. 1, pp. 9–34, 2015. Data files are available for confirmation. http://www.scientificexploration.org/docs/29/jse_29_1_OshopandFoss.pdf
You may also want to look at my videoblog discussing the relationship between astrology and science. https://www.youtube.com/c/JutkasAstrologyandScience
Prediction of one of the points made in the next NYT hit piece about ACX: "Scott Alexander gives platform to astrologers" (90%).
But not to professional gamblers? I am a bit stunned by the amount of hostility here. Is this necessary? Let me explain my position. I am aware of the risk Scott Alexander took by addressing me and am grateful for his open mind. Other well-known cognitive scientists and AI researchers did the same on Twitter and Reddit but I am not going to name them to preserve their professional reputation. Look at what I am saying while science claims to have an open mind (?). This is why using ML is a good idea: to put astrological predictions on a relatively unbiased platform with large public data sets.
I think astrology should be put to the same standards as the social sciences. No more no less. Same predictive values. We are dealing with human beings here and predictions have thousands of variables; in addition, we astrologers don't have a unified body of ever-increasing knowledge behind us because of the disruption of the Enlightenment. Since then astrologers needed to work outside of the Academy. Now the Academy is in shambles and because of the Internet and an increasingly skills-based job market we are returning to a more personal way of accessing knowledge. Perhaps our time has come.
With help of ML we can put thousands of years of multicontinental astrological traditions to test. Yes, there are charlatans...there are none in science? Yes, some of our predictions were incorrect...none turned out to be so in science? But as far as I am concerned there is absolutely no competition here. I love science but not scientism. Very grateful to be alive in these exciting times.
In my view, the only two aims of astrology is to help clients overcome their challenges and make useful predictions about the world around us. I have addressed the second above. As for the first claim, I believe that we are all born with a horoscope which is a map of the planets at the time of our birth. This map is a symbolic representation: a blueprint or a bit like a negative of a photo. You can develop this negative well or badly but you cannot change the negative. A personal consultation with an astrologer is similar to a session with an analyst helping to develop the potential of the client --embedded in that horoscope to the maximum.
This is my personal view: a rather fatalistic approach giving very significant weight to the givens in the nature/nurture balance and quite aligned with our contemporary views on the role of genetics in human development.
Finally, for fun, and this is purely speculative if there is any validity to the simulation theory and in the unlikely event that we ever discovered that it would not surprise me to find horoscopes to be part of the digital construct.
I am happy to offer a private consultation online to any one of you to decide for yourselves. My only request is that you provide your correct birth data. I don't need to know your name and prefer not to see your face. Of course, if you find the consultation useless you are welcome to a full refund and publishing your opinion in any forum you desire. Jutkasastrologyandscience@gmail.com
Thank you for posting this. Regardless of how right/wrong you are, defending yourself with data and not falling for the baity comments is commendable.
Agreed. I do have to point something out, though. Lets assume that you're right and there is astrological research that makes valuable predictions. But then you included two examples that are really underwhelming because the claimed results do not seem all that useful. If that's really the best you have, that's not much.
You are hoping astrology can be held to the same standards as social sciences. I'm not in social sciences, so maybe the similarity is closer than I think, but it's really hard for me to see the value of predicting typos in Amazon reviews or number of celebrities' Twitter followers. If I was tasked with reviewing a technical article that predicted these things based on something I believed could predict them (such as native language, geographical location, education level, type of device used for input), I personally would not recommend publication, because I would fail to see how researching this has value. I'd be surprised that someone funded it (or not, there's been a lot of funding for all kinds of useless or ridiculous projects).
If you want astrology to be held to the same standards as social sciences, you're going to need stronger examples than what you cited.
Here is another open access astrology-based ML predictor for alcoholism vs teetotalism included below. Please feel free to try it.
Meanwhile, the examples above are not trivial at all when considering the following:
Trying to publish on the subject of astrology in a scientific journal is quasi-impossible. The prejudice is insurmountable. These particular studies were done because the subjects are familiar to the public and the author needed a huge and openly accessible free data base. Amazon provides this from the time the first reviews were written. Ms. Oshop also needed the kind of results that are not subject to interpretation, in plain words, results that are not easy to explain away. Finally, for most astrological predictors to be precise you need the exact time of birth (not just day). In the case of celebrities, they are relatively easy to attain and have the predictions publicly verified.
I agree with you about the very relative importance of Twitter followers, however, to be honest for a moment here, is this the general impression you get of AI researchers, cognitive scientists, engineers, investors, science-podcasters, etc. working either in or outside of Silicon Valley?
Admittedly, I myself have probably chosen a different subject but I am not a programmer and understand that a lot of thinking goes into how to formulate a study that fulfills all requirements on such a taboo subject. Maybe there was also a consideration of public appeal.
However, ultimately, I don't think it is relevant whether the original study is about Amazon misspellings or Twitter followers as long as it demonstrates that astronomy of day predicts with stronger values than other traditional methods. (The Amazon study does point to a very strong correlation with Mercury retrogrades, for example). Once it works you can use it for anything you like including all kinds of useful Covid related predictions.
The hard part is to have someone look at it, have the courage to say maybe there is something here (if there is). I would be so grateful if an expert looked at the study on its own merit (methodology, results, etc.) because I am not a programmer myself but then we would get somewhere.
OK,here is the other example you can try for yourself. It predicts for alcoholism vs teetotalism.
Easy Machine Learning Using Astrology, http://www.instructables.com/id/Easy-Machine-Learning-Using-Astrology/
It's open access and was written for people with no knowledge of programming. You can immediately see all the problems I have described above hence I have not included it in my previous post. Also the data base is way too small which is why I would be grateful if anyone tried it for themselves using exact birth data of people you know well and therefore can easily verify the results. An acquaintance who has been betting on horses using ML for years just repeated it with his own spreadsheet and found it reliable.
Thank you
I'm not a scientist myself (yet), I would love it if you shared your opinion about the papers bellow.
Without understanding everything about https://ijsrm.in/index.php/ijsrm/article/view/2040/1719, here is my take on its conclusion:
"Future misspelling rates in Amazon reviews were successfully predicted using only basic astronomy data in a neural network".
I imagine misspellings correlate very highly with the number of reviews written by non-native English speakers. I'm also quite sure that the number of non-native English speakers posting reviews online depends on the time of the day. It is a lot more probable that you write a review at 4:00 PM rather than 4:00 AM. It would be trivial for a ML algorithm (or a classical algorithm) to calculate which countries were awake given so much data about positions of different planets relative to us. So it seems to me that we are looking at an fancy "what time is it" calculator. Yet there is so much I don't understand because of my lack of education that I may be completely misreading it.
Hello! The reviews were averaged across each day (with 5296 days included) and were English-only.
And here is a correlogram of these average daily misspellings within Mercury retrograde periods https://www.ayurastro.com/uploads/1/1/0/9/11099376/correlelogram_orig.png https://en.wikipedia.org/wiki/Correlogram
For a bit more on the background to the correlogram see https://www.ayurastro.com/articles/ridiculously-good-results-again-for-the-amazon-misspelling-rates-data-using-bigmls-deepnet#/
I still think this way of giving and judging predictions, using confidence levels, is fundamentally flawed.
Here, let me make some predictions with confidence levels:
1) The sun will rise in the east on February 24th (95%)
2) The sun will rise in the east on February 25th (95%)
3) The sun will rise in the east on February 26th (95%)
4) The sun will rise in the west on February 27th (95%)
5) The sun will rise in the east on February 28th (95%)
6) The sun will rise in the east on March 1st (95%)
...
20) The sun will rise in the east on March 15th (95%)
If we evaluate my predictions a month from now, I think you will find that out of 20 predictions I made with 95% confidence, I got 19 correct. Perfectly calibrated! Clearly I'm the greatest predictor on earth!
I think the only good way to score predictions is to compare them to other people. If a hundred people predict A, you predict B, and you're right, that's great. If 80 people predict X, 20 predict Y and you also predict Y, that's pretty good. If 5 people predict N and you and 95 others predict M, then that's not impressive at all, even if you're right.
"I think the only good way to score predictions is to compare them to other people."
But isn't that what he did in some form or fashion here? He even says that it's not so much about absolute score as relative position, "Wringing an absolute judgment from a scoring rule like this is hard, but getting relative standings is more straightforward. I did better than Bucky, and Zvi did better than me."
Yeah my point was more a general point about the way predictions are often treated on this blog and the wider community. Something I've been meaning to say for a while but never got around to.
What Scott is doing here is a step up from how he did it in previous years. But 3 is a very small sample size, and he scores based on self-proclaimed confidence-intervals instead of based on 'distance from the norm'. Zvi scores more points for his mail-in ballots prediction than his China prediction, despite the latter being, to me, more impressive. Predicting someone that everybody else is also predicting is not impressive.
Do you have an idea for a more practical way of how he could do this?
Well the easiest way would be to make a long list of yes/no binary predictions, and then let ACT readers fill those out. Make it a little survey.
That only makes you look like a good predictor because you're gaming the system, picking a bunch of guaranteed events and deciding what rate you want to get them right. It is really obvious that you're doing this because we all know the sun is definitely going to come up tomorrow, so no one is going to think you're a great predictor. And if you can get perfect calibration while listing events that everyone doesn't think are obviously guaranteed, then it sounds like you're making valuable predictions.
Yes. Exactly. That's my point. Predictions should be measured against others predictors, not against self-set confidence intervals.
Typically, the way predictive model scoring is done in industry is to compare against a null predictor of some sort. For the sun coming up, the null predictor would be yes every day for the next five billion years, so if you predict yes, your accuracy premium above the null predictor is nothing.
Scott is doing something *sort of* like that by setting 50/50 picks to be scored at 0 if they end up being true, but that isn't quite the right way to do it. I'm not totally sure how you define a baseline predictor for arbitrary binary events, though. Random prediction would work if you were actually making binary predictions, but throwing in the confidence estimations messes things up, and it only really works (and justifies 50/50 being 0) if these are truly toss ups.
But to better demonstrate why I think what Scott is doing is degenerate (agreeing with you, as it stands), take something like sports. The top team in the NBA right now has a 80% winning percentage. If I just predict they win every remaining game this season and say I'm 80% confident of that, I'm likely to end up scoring very high by Scott's scoring system, yet I have effectively just made a null prediction and should score 0.
That said, he's clearly doing the right thing with the over/under type bets, though I think in that case you have to weight the scoring by how close you ended up being and not just all or nothing.
The absolute score is not very meaningful, for exactly the reason your example illustrates: it doesn't have any way of accounting for the "difficulty" of the predictions. The point of the scoring rule is that if you are trying to maximize your score, you will use your true beliefs.
Are you suggesting that he shouldn't use confidence levels at all - that he should make deterministic predictions?
I mean, in terms of how useful someone's predictions are, I can agree a term for surety is useful - obviously, given someone whose predictions are well-calibrated, the surer they are of things within that constraint the better. However, I think somebody who's perfectly-calibrated but whose predictions are evenly distributed between 50% and 100% is probably still more useful than someone who always predicts deterministically but gets it wrong 5% of the time - if nothing else, you know which of their predictions you can rely on (if those predictions are all minor then that's its own problem, but that's more of an argument for some sort of weighting based on importance along with the aforementioned surety term than for comparing them to others).
Also, how would you score wrong predictions in your model? If I predict something bizarre that nobody else did and I'm wrong, is that better or worse than if I predict the same wrong thing everyone else did? (My understanding is that strategy-invariance would require "worse", but hey, it's your model.)
Each week I share in the Metaculus discord a list of the top comments. If people are interested, I will share that here too. (Here is last week's:)
https://metaculusextras.com/top_comments?start_date=2021-02-15
Last week's top comments:
- More discussion of Elon's prediction: https://www.metaculus.com/questions/6498/will-sn10-land/#comment-55669
- Assessment of Biden's chances of running: https://www.metaculus.com/questions/6438/will-joe-biden-run-for-reelection/#comment-55732
- Thoughts on Alcubierre drive in the wider context of theoretical physics: https://www.metaculus.com/questions/6558/working-alcubierre-drive-before-2100/#comment-55878
About the Alcubierre drive solution, I think people are underestimating just how low physicists' priors are on things like superluminal signaling. Describing violating energy conditions as a "technical challenge", like the metaculus question does, doesn't really capture the problem there. Also, any sort of superluminal travel that works at all like you'd want it to (by connecting two distant points in otherwise flat spacetime) necessarily implies time travel. If I had to guess, phrasing the question like "will a working time machine be demonstrated before 2100" might elicit a lower prediction than the question as it's worded now.
I never understood the argument for why FTL necessarily implies time travel. Suppose I somehow invent a teleporter that lets you step into it, disappear in a puff of smoke and appear at the target destination t seconds later where t is the distance traveled divided by ten times the speed of light. I cannot use this machine to accomplish anything a normal person would recognize as time travel, no killing Hitler or shaking hands with my future self. Why is this a problem?
Consider me unconvinced. I've parsed quickly the Wikipedia page describing how FTL signaling could lead two people moving relative to each other to hear a response to their signal before or was emitted, and it seems to rely on using relativity formulation of Lorentz transform, which FTL itself makes wrong. FTL would mean you can mesure absolute velocity, i. e there is a prefered frame of reference (for FTL stuff, not for the non -FTL physics). So no, FTL will not lead to time travel in the classical sense. It will just break special relativity and reintroduce the aether. No big deal, except no FTL signal has ever been observed
Oups, i read Wikipedia to fast and missed the last sentence : they explicitly mention that if you allow for a preferred frame (for FTL) causality is not violated. So i was right, FTL does not violate causality, it re-introduce the aether...and I stand it is no big deal, cause the insistence on being unable to define a special frame of reference is imho the weak point of special relativity. Background radiation already put it in jeopardy...
Aether is a sufficient but not necessary chronology protection for FTL. FTL can be consistent with causality without a special frame as long as there's some physical mechanism that breaks the FTL if a causal loop threatens to form.
(I've seen people propose that quantum vacuum effects would collapse any wormhole or series of wormholes at the instant they permitted a closed causal loop. I'm not sure whether a similar mechanism has been proposed for the Alcubierre metric.)
Note that I'm not saying FTL is necessarily possible (or, for that matter, that causality is necessarily true; time travel doesn't *have* to result in paradoxes, although consistency protection without causality gets bizarre fast).
Sorry, my answer is wrong (there's some wiggle room with the exact definitions of "t seconds later" though). For a reasonable definition of faster-than-light, it's more complicated than warping out and then flying back.
If your definition of warp speed is invariant for moving observers, then there will be a way to travel back in time: You'd need to be able to warp twice, and the second time, you need sufficient velocity relative to your original position. The trick is that faster-than-light travel for one observer will appear to be backwards-in-time travel for another observer with high relative velocity.
You can get around this by saying that warp speed is always defined in terms of Earth's frame of reference. You lose invariance of physics (and therefore just about all our theories of fundamental physics!), but you can regain causality.
I would not choose earth frame, but some absolute frame. Exactly back to the aether, the version whose Lorentz transformation tried to reconcile with Michelson&Morley experiment. Personally I like this framework, just before Einstein went a step further: as no mechanical nor electromagnetic phenomenon could measure motion against the fixed frame, why not get rid of the fixed frame and use a formulation that make it dissapear from the mathematical desciption? It was a good idea, as invariance also holds for non-electromagnetic phenomenons...
However, any FTL information transfer would mean that relativity is not absolute (for this you need a single max velocity), so back to a fixed frame (that you can even define using your FTL). I don't think we will see FTL, not because it would create time-travel paradox, but because c is really the max speed for anything we have observed until now. We have observed some effects very different from mechanics and electromagnetism that produced special relativity: gravity waves, electroweak, nuclear....so FTL would be a new new thing, and I find it unlikely that new new thing break an invariance that was verified by multiple new things.
Still, I am still fond of pre-einstein absolute frame: maybe because I make less mistakes using frame of alice, frame of bob, and an absolute frame with an absolute time than just alice and bob, maybe because doppler shift of cosmic background radiation looks eerily close to a physical absolute frame...
https://en.wikipedia.org/wiki/Tachyonic_antitelephone
FTL travel like you describe would make shaking your own hand possible. The idea is that which events can be said to occur "t seconds later" actually depends on your velocity. Say you and your target location start off at rest relative to each other. According to special relativity, changing your velocity so that you are now moving *away* from your target effectively pushes your target's timeline forward in time so that some events that you previously would have considered to be a part of your target's past are now, from your new perspective, in your target's future. Move away from your target fast enough (>c/10) and a sudden decision to deploy your 10c apparatus will send you to what you previously regarded as the past of your target. To shake your own hand, you just need to turn around and do the same procedure back to your starting point.
> Also, will Bitcoin outperform the US stock market over the next five years, at 51%. I started out thinking - of course it's 50-50! By the efficient market hypothesis, if any asset was obviously going to do better than another, people would change the price until it wasn't. But on second thought that's wrong - stocks have a higher than 50% chance of beating treasuries over the same period because of a risk premium. Maybe there's no intuitive way to think about this, you have to have opinions on the underlying fundamentals, and it's only 51% by coincidence?
Yeah, this is wrong, and I was also tripped up by this a while ago. The efficient market hypothesis only tells you that E[X] = X. But, e.g., if Bitcoin could only either go up 3x or crash to 0, then the efficient market hypothesis would tell you that p * 3 * Initial capital + (1-p)*0 = Initial capital => p = 1/3 = 33.3% of going up and 66.6% of going down.
It looks like you're still collecting play-money prediction markets but haven't mentioned The Foresight Exchange, one of the oldest. (I tried to post this last week.)
Perot did not win 5+% of the vote as a third-party candidate in 1992 and 1996. He won 8% of the vote as a third-party candidate in 1996, after having won 18% of the popular vote as an *independent* candidate in 1992.
It's an important distinction, because running for President as an independent candidate implies that you're trying to pull the Gabriel-over-the-White-House trick, riding into Washington on a white horse and sweeping away the corruption with the awesome power of the presidency. If you envision *not* being a dictator and actually sharing power with other people, then the first step is almost certainly going to be finding a bunch of other people (particularly wannabe-Congressman type people) with similar goals and getting organized.
Metaculus is scoring the prediction as positive if either a third-party or independent candidate reaches 5%, so at least they're aware of the distinction. But they don't seem to think it is an important distinction, and I think it's important to push back against that.
I think to most people, running as an independent is just a signal that you don't like political factions. You can refuse to identify with a faction while still respecting separation of powers and checks and balances, working with existing factions, and just generally not being a dictator.
I don't really think it's a meaningful distinction for the purposes of the prediction.
My criticism was going to be that only looking at the last eight (at the time of prediction) elections was cherry picking.
BUT it turns out the a third party or independent candidate has received >5% of the popular vote in 11/58 (19%) presidential elections (again, at time of prediction), which isn't *that* far off from 2/8 (25%).
[insert more you know gif]
What exactly is the distinction? If I win the presidency after running as an independent, is there any difference in how I am allowed to exercise my new Presidential powers compared to if I had won as a third party?
To a first approximation, the president has zero power except insofar as his allies in congress pass budgets and legislation to support his plans, and his staff in the White House competently and loyally executes those plans. Assembling a legislative coalition and a capable executive staff is close enough to establishing (or co-opting) a political party as makes no difference. If instead you think you're going to walk into the White House as an outsider, appoint a few of your outsider friends to the cabinet, and actually accomplish anything, then no, that's not going to happen.
None of that is a distinction. Is there a law preventing independents from forming coalitions? Will everyone in Washington refuse to talk to you if you declared as third party? What can an independent president do that a third party can't?
There's no law preventing someone from winning a (hot) war without an army, or manufacturing a million automobiles without a corporation. In theory, you could own a car factory as your personal property, hire thousands of people to work for you personally and tell them all to go run your personal factory, etc. But the organization that can build a million cars is a corporation whether you call it that or not. The force that can win a war is an army whether you call it that or not. And the coalition that can govern a nation is a party whether you call it that or not.
There *are* e.g, ballot access laws that make it significantly harder for you to do this if you don't call your coalition a "party" and fill out the paperwork as such.
And, none of the people who have run for the presidency as non-partisan independents in the past century or so, have made any real attempt to build such an organization. It's not something you can do in the two months between election and inauguration. Perot tried to do it after his 1992 presidential bid, but appears to have lost interest and phoned it in come 1996. "Independent" presidential candidates aren't people who have some perverse aversion to using the 'P' word to describe their thing that is a party in all but name. They're people who expect to play the Gabriel-over-the-White-House game.
"And the coalition that can govern a nation is a party whether you call it that or not."
Counterpoint: In Canada we have more than two parties that routinely win seats, meaning we often end up with a minority government that has to make deals with the other parties to get anything done. Sometimes you get a coalition that lasts a few years of party A and B working together to do all the things they agree on, but sometimes you get party A teaming up with B to pass environmental regulations one week, then teaming up with C to bust unions the next, over B's protests. Sometimes the MPs even vote against party lines which really muddles things.
If all governing coalitions are parties, the shifting nature of our coalitions either has new parties routinely popping in and out of existence, or people are leaving and rejoining one superparty on a weekly basis depending on what's up for a vote. This seems like an odd and not especially useful definition of "party".
But why is it *Party* A & B working together to do these things, rather than Party A + Party B + that guy who used to be part of party B but realized he could negotiate for things more specifically tailored to his constituent's needs if he went independent? And then next year just a hundred seventy independent legislators each seeking their personal optimum outcome? Legislators give up the ability to seek some things valuable to themselves and their constituents when they join a party. Why would they do that, if "party" is a silly meaningless null-word?
"Party", means in this context "organization actually capable of entering into the sort of negotiations you describe". Hundreds or even dozens of independent legislators, really can't do that. That's why they form parties, and that's why "party" is a useful word.
A pet peeve: "general consensus": If it's a consensus, it's general. "Consensus" is accurate and has less redundancy.
I think there's some value in distinguishing between consensus among those who are most informed ("expert consensus") vs among people in general. Maybe it's a bit redundant but it is not without information content.
"<i>9. Will China’s reach 100,000 official cases?</i>
China's official case count on 12/31/20 was 95,963, so false."
That was so close. Should there be a scoring system which adjusts for predictions which are just barely true or just barely false?
A probability distribution over case counts, rather than a fixed-point, would have given most of the due credit here. The only downside is that a well-calibrated probability distribution requires much more thought.
Having lived in China, I regard their numbers as suspicious. Known to me is that if you're in the West and you die of an MI while COVID-positive you get scored a "COVID Death" whereas in China you're a "Heart disease Death."
Moreover in 2020 China didn't lock down Wuhan until After the annual Great Migration for Spring Festival had started. Many of my fellow ExPats had returned to their home countries, with varying local rigor of quarantine. ExPats or none, an annual migration of a billion people to and from the largest cities on Earth is not conducive to a quarantine.
My opinion based on my limited knowledge is that China's death and disease toll, were it counted the way we're counting, would be similar to India's given the similar living conditions of their poor and minority populations.
But of course, as a Westerner, I Just Don't Understand China—so say my coworkers—so it's entirely possible that regarding this ONE thing the Chinese government is NOT withholding embarrassing data that would cause them to "lose face" internationally.
They were well aware of this issue, and phrased their prediction carefully so that it is about the *official* case count.
How useful is the prediction when you may be scoring in (large) part the extent to which China is willing to lie?
Don't you think it's useful to have some thoughts about how much the Chinese government is willing to lie, and for those thoughts to be accurate?
But then you'd want to score a lie where it's clear what the truth is, so you can see the deviation. And you'd want to present it as a prediction on that, rather than on COVID.
Neither are true here.
If you have all three measurements, then sure, you want to test your prediction on all three. But when you know that all you have is the combined noisy measurement, testing your prediction on that is better than testing nothing at all, or testing your prediction of the thing that you claim without measurement.
"similar to India's given the similar living conditions of their poor and minority populations"
Is this true?
In my experience, going only places where Westerners are reasonably allowed to go is that China's poor whom I observed were living in unthinkable squalor. You won't see that written about in Xinhua, but like Shanghai Pride, also not reported by Xinhua, nonetheless exists. Colleagues who have been assigned to both countries tell me that China probably wins on sanitation and violent crime.
Yeah; China has many poor regions; it is about as rich a country as Thailand. But both Thailand and China have been foresighted about quashing COVID outbreaks. The Hebei outbreak is over, and the Thai outbreak is becoming increasingly confined to Samut Sakhon.
"would be similar to India's given the similar living conditions of their poor and minority populations"
Impossible. China has been far more eagle-eyed about stamping out infections than India. Seroprevalence surveys have been consistent with the official death toll in Wuhan.
In grad school I took a course on probabilistic risk where our tests were graded using log odds scoring. Instead of picking A, B, C or D, you wrote in your probabilistic confidence in each choice. For example, if you’re somewhat confident in A, write in 60%, and then distribute the remaining 40% of weight across the other three choices.
Because of the log scoring rule, being very confident and wrong would result in a massive penalty. There was one star student in the class who blew the curve for all of us by putting down 90%+ confidence in his correct answers because he knew what he was doing and, furthermore, knew that he knew what he was doing.
This method was really hard and I got a pretty mediocre grade in the course, but I’ve always thought it was a good idea. You really learn about your own understanding of the material this way. It’s much easier to notice gaps in your own understanding when, instead of being given partial credit, you’re actively penalized for undue confidence.
"But the soul is still a mantic thing; amid the market's din,
List the ominous stern whisper from the Delphic cave within,—
'They enslave their children's children who make compromise with sin.'"
You know, that really works for me.
A 3rd party POTUS candidate reaching 5% is not a particularly high bar nor, during my lifetime at least, a highly unusual event. It happened in 1996, 1992, 1980, and 1968. Three of the last 11 presidential elections and four of the last 14. So even before considering anything specific about 2024, I'd take the "yes" side on 15% odds.
On the third-party candidate: I think that the naive assumption of "Last 3 of 8" making the 15% bet worthwhile is a bit simplistic of an analysis. The first thing to consider is that, for such a small sample size, if it's possible to get more granularity, one should. And we certainly can. The political environments of those 3 third-party successes, so to speak, were not nearly as polarized as things currently are. Unless you think there will be a de-stressing of tensions (Possible) or a fracturing of parties (Very unlikely, given the two-party system's strength) I think it's closer to 1%. If there's an easing of tensions, call it 20%, very generously. Take this as me predicting against this.
Incidentally, it feels to me that there is some serious ambiguity on the "second wave in fall" question. Some would say the second wave was in the summer, and the later one was a third wave. Some would say that third wave was really in the winter, because it peaked in December and January most places. Some would say that although the summer wave seemed distinct from the first wave, it was really just a delayed first wave that hit Florida, Texas, and Southern California as soon as the initial restrictions were eased. Some would say that in the northern plains region, the wave that started building in August/September was actually just their first wave. Though I think that New Mexico and El Paso were clearly in a second local wave by October/November.
But it's really hard counting waves.
"highest death toll per capita (although all the cities that had higher death tolls per capita were small"
Which cities had higher death tolls per capita?
Here is an interesting paradoxical problem with this type of prediction self-scoring: suppose you are especially adept at thinking up prediction questions that are "hard" precisely because they successfully "cleave (potential future) reality in half". For example, your question 9 asks if China will have 100k deaths. In reality it had 96k deaths, so the threshold in this question is a pretty good guess!
If you were insightful enough to keenly set the thresholds for all your questions in such a way, you would frustratingly find that there is no way to earn a positive score at your own game (or even have much fun playing it), since an optimal Bayesian agent should predict 50-50 for all questions.
Re: "will Bitcoin outperform the US stock market over the next five years"
Consider my new cryptocurrency, Gamblecoin. On Jan 1, 2022 I will roll a die, and if that die comes up 6 it activates a smart contract that will buy anyone's Gamblecoins for $1 USD each, otherwise nothing happens. The odds of Gamblecoin beating the market are one in six, and this is not an efficient market violation because if it does beat the market, it beats it by a lot.
On a long enough timescale, Bitcoin should either become some kind of enormous global currency worth way more than it currently is, or become basically worthless as we find something else to serve the "anonymous secure global currency" role (doesn't have to be crypto, maybe the banks get their act together and come up with an efficient way to move USD) and Bitcoin no longer has a reason to go to the moon. The odds of Bitcoin becoming an enormous global currency are probably not 50/50, so there is some time period over which prediction markets shouldn't say it's 50/50 to beat the market.
That being said, there's a decent chance that Bitcoin is still in the speculative bubble phase and we won't know whether it becomes the massive global currency until far in the future. "Does Bitcoin beat the stock market over the next five days" should be 50/50 on the efficient market grounds you mention because the next five days are (very probably) just going to be speculative trading rather than a resolution to the question of whether Gamblecoin has value or not.
Vaguely on topic: somewhere in a thread on SSC or a related site, I predicted 1 million or more excess deaths in the US during covid, explicitly including deaths due to untreated non-covid conditions, despair due to unemployment, etc.
A month or two after that, US covid deaths were still below 1/4 million, and the rate was dropping, and I felt like an idiot - but I'd made the prediction, as my first attempt at measuring my prediction results.
When I checked today, the number of explicitly covid deaths reported in the US is over 1/2 million (501,117), and my confidence in my prediction is much higher.
Excess deaths - i.e. number of deaths in each 12 month period compared to the last 12 months without covid - aren't tracked as prominently, or reported as swiftly, so I can't yet check how close we've come so far.
I'm reposting mostly to continue to hold myself accountable.
Regarding bitcoin vs. stock market:
I see a number of people modeling bitcoin outcomes as binary. That may be reasonable, but just about every other financial asset is very well modeled as a random walk with drift (expected return). If we do that, stocks have about 15% annual volatility over long horizons, and bitcoin has... I actually haven't measured it but it's probably 5-10x that. Expected returns on stocks is probably in the ballpark of 4%, looking at international markets (US has done better, but using its returns to predict its future returns is probably worse than using the average international estimate). Over a 5 year period, the volatility of bitcoin is so much higher than stocks that the average return on stocks barely matters. So the random variable (bitcoin 5 year returns - stock 5 year returns) has a negative mean (if we assume bitcoin doesn't have a positive risk premium), but its volatility is so high relative to that mean that the probability of it being greater than 0 is pretty close to 50%.
Careful: if you assume a lognormal distribution of returns (which isn't empirically accurate, but it's usually the default approximation), higher volatility doesn't always translate to a higher chance to cross above a barrier. As volatility gets high, most probability mass concentrates around 0 in a lognormal distribution.
That’s a good point. I was being lazy about normal vs lognormal, and at these high volatilities it matters a lot. So I take back my comment - it actually may be reasonable given this level of volatility to say that it’s either zero with some high probability (or close enough to it that you clearly underperform equities) or much higher than it is now, and the probabilities around those two possibilities are not necessarily 50/50.
Have you made the 2021 predictions? Or will you continue the cycle and make them in April?
>16. Will there be a general consensus that summer made coronavirus significantly less dangerous?
Maybe this is true, but I have my doubts. I found this posted on Good Judgment Open:
"Given the lack of immunity to SARS-CoV-2 across the world, if there is an effect of temperature and humidity on transmission, it may not be as apparent as with other respiratory viruses for which there is at least some preexisting partial immunity. It is useful to note that pandemic influenza strains have not exhibited the typical seasonal pattern of endemic/epidemic strains. There have been 10 influenza pandemics in the past 250-plus years—two started in the northern hemisphere winter, three in the spring, two in the summer, and three in the fall. All had a peak second wave approximately 6 months after emergence of the virus in the human population, regardless of when the initial introduction occurred."
https://www.nap.edu/read/25771/chapter/1
I will add that the Winter Is Coming! narrative probably seemed like convenient messaging to some of the public experts and encouraged the perception that there was something special about winter that made the 2nd wave so severe. It fed into public intuitions about winter being worse for the cold and flu season.
But if we go with the above notion that the 2nd wave was pretty much bound to happen regardless of the season, it seems more a matter of luck that the 2nd wave did coincide with winter. Of course, that doesn't answer the question of whether the 2nd wave coming in winter did make it significantly worse than it would have been otherwise.
My intuition is that 2nd waves in pandemics are likely to be worse. Unlike the 1st wave, there is no "patient zero" in a given city who orbits within a few social circles. As the 1st wave subsides and R0 waffles from >1 to <1 to ~ 1 in a given local, the disease spreads gradually to people in all walks of life, orbiting all social circles, in all communities, so as the 2nd wave starts due to say a general relaxation in social behavior, we now have a thousand "patient zeros" in a thousand locations, a thousand sparks able to ignite the kindling in a thousand more locations, bringing the starter wood to a hotter temperature and ultimately burning the forest with more ferocity than the fire could six months prior.
I agree. I was actually surprised by the sentence
"Most papers I read now agree that coronavirus is a seasonal disease and that it's not a coincidence that cases went down in summer and up in winter. "
Because I also tried to hobby-research this question just a few days ago, and came to exactly the opposite conclusion. The papers that convinced me most found a non-zero, but pretty minor effect, like a ~10% increase in the R-value if temperatures rise by 20°, which is roughly the difference between European winter and summer. I also looked anecdotally into the numbers, and most European countries had a quite strong and consistent exponential growth during all of summer.
It seems that you can find a correlation between "winter" and "high numbers" in a lot of countries. As Jack Wilson points out, this is probably coincidence and tricks intuition of many people. Actually, this connection doesn't make any sense. It should be between "winter" and "high growth rate", and *that* connection does not seem to exist. It does exist for influenca, which starts growing around September/October and doesn't reach high numbers before January/February, but corona doesn't seem to follow this pattern.
I might add that I find it possible that covid-19 does become seasonal in the future. Once immunity in the population keeps it just below the critical value, then a 10% increase may become the push that brings in into supercritical regime. But I think we are still far away from this point.
These markets seem to be modeled on, well, markets, but the function they aim to fulfill seems to be closer to that of a consulting firm. Betting/investment markets, at least in the US, come with a prohibitive degree of red tape.
So, bizarre question: would it be possible to create a prediction "market" that functioned like "the Uber of consulting firms"? Clients contract with the firm for a given amount with a set of things they want predictions on. "Independent contractors" can buy a percentage in the contract and submit their prediction(s) for the associated questions. Payout is a function of accuracy and percentage of buy-in.
1) Could that work?
2) Would it be legally less difficult than prediction markets as currently constituted?
Have a look at The Good Judgement Project - is that the kind of thing you mean?
Yes, along those lines, but opened up on the prediction end.
Re Third party 5% share: Seems like the easiest path would be a Trump Patriot Party-esque splinter. This'd require Trump to rerun (or find a charismatic substitute) and for the GOP to resist hard enough to create an election-losing vote split over it, which seems iffy, but certainly raises the chance beyond the background chance in the previous elections.
Last year, I built a site that was based on predicting the future. (Don't worry, it's dead, this is not an advert)
The idea was to make it a more general, lower-entry prediction market. Instead of people betting money, they would gain or lose reputation. The site itself would function like a geopolitics focused subreddit, except that people with good reputation have much stronger upvotes. Therefore, the smart people would rise to the top much faster, and if a strong expert made a comment a few hours too late, it would not be buried. Second, any unproductive groupthink or echo chamber effects would be broken down quite fast - the people that succumbed to this sort of stuff would lose reputation, and the people that didn't would rise up, relatively speaking.
And of course, the reputation of every poster would be visible next to their comment for bragging rights.
In the end, I couldn't figure out how to reach a critical mass of users and took it down; here is a mock-up I used that shows how it worked, if anyone is interested.
https://app.moqups.com/7XPcEzrtnD/view/page/ab67a74d4?ui=0
Gladly. Please check the following two publications demonstrating that planetary positions correlate with phenomena with high r2 values written by Renay Oshop my project partner.
"Astronomy of the day is more effective than seasonal decomposition in modeling and predicting the rates of Amazon review misspellings," International Journal of Scientific Research and Management, Feb 2019, Vol. 7, 2, DOI: 10.18535/ijsrm/v7i2.aa01, https://ijsrm.in/index.php/ijsrm/article/view/2040/1719
and
"Twitter Followers Biased to Astrological Charts of Celebrities," Journal of Scientific Exploration, Vol. 29, No. 1, pp. 9–34, 2015. Data files are available for confirmation. http://www.scientificexploration.org/docs/29/jse_29_1_OshopandFoss.pdf
You may also want to look at my videoblog discussing the relationship between astrology and science. https://www.youtube.com/c/JutkasAstrologyandScience
Those papers were hilarious, thank you.
On third party vote share: Gary Johnson got 3.27% of the popular vote in 2016. I think if you re-ran the tape a few times he might have crossed 5%. Indeed, FiveThirtyEight's final forecast (https://projects.fivethirtyeight.com/2016-election-forecast/) had him at 5.0%.
Nobody you personally know outside of patients got Covid?! Wow, either your social circle is very small or very cautious.
Talking to some people on Twitter suddenly reminds me how, as a Gen-X er, I'm still bitter about how much the government in the form of Surgeon General Dr. Koop scare mongered straight people about AIDS. They wanted you to believe circa 1990 that all oral sex was going to give you AIDS if you didn't wear a dental dam.
I'm not saying that mattered for most people, but for many of us on the margin I think it was a big deal. Sex is awkward when you are young and the fear of AIDS made it much more awkward. Not much is written about that these days, but when I think back it was a huge deal at the time.
Agree. Aids was so scary back then with magazine PSAs showing an attractive/available woman an then telling you on the next page that she actually has aids and you could die if you had sex with her, - "You're not only sleeping with her, your sleeping with everyone she's ever slept with any everyone they've slept with too." Also a mirror page showing you a "typical aids victim". While I was losing my virginity, I was thinking about Aids infection - "I hope to god this condom doesn't break or any fluids leak round the side, I know she's only had one boyfriend before but how many people has he slept with...??? I don't want to contract HIV and die of aids. Fuuuuuuu". Not conducive to a fun relaxed first-time shag.
How do we know if we over- or under-reacted? It seems like a lot of people are locked in to defending their views on what the right level of reaction was.
By the way, here's a question for helping come up with an objective assessment of how bad covid was. The Spanish flu of 1918 hit men between 20 and 40 the hardest, thus leaving a lot of widows and orphans. With 500,000 Americans now having died from covid, how many orphans (under 18 years old) did this toll create?
Just wanna say it's so strange that neither you nor your friends got covid-19. I am russian and it highlights again, how different is our risk tolerance, comparing to the rest of the world. I remember one link you posted earlier, where some guys try to calculate the risk of some some interactions, going from the assumption that the reasonable risk of getting covid per year should be 1%. For me it's ridiculous. I would say 70% chance of getting covid per year is ok, maybe 40% if you are extra-cautious.
I got covid (asymptomatically), my wife got covid, my sister and her husband got covid, my wife's cousin and all her family got covid, my best friend got covid (asymptomatically) and his mother got covids, some other friends also got covid. (No deaths, one hospitalization).
And Russia has so few restrictions. After spending five months last year in Chile in a lockdown I enjoy the freedom to leave my house tremendously.
All response are a matter of conjecture. Posit if you will. Nobody I have seen in the public format is a prophet; it's a good thing to give and take in an opinion which can be refined to it's essence without another opinion.
I made predictions last year, inspired by Scott. I got 7/8 95% ones, 15/16 90% ones, 10/13 80% ones, 6/12 70% ones, 8/12 60% ones, and 2/12 50% ones. It seems like my largest error was in the 50% ones, strangely enough, I think due to my tendency to throw a lot of low-probability things with affirmative phrasing ("this WILL happen") there. As a fun fact, the 95% one that I got wrong was "I will still regularly be reading SSC at the end of the year: 95%"; this was incorrect for reasons entirely different than I imagined.
You've inspired me to evaluate my own predictions (mostly relevant to Georgia-the-country) from last April - https://peripateticpedagogue.wordpress.com/2021/02/24/evaluating-past-predictions/
In retrospect, it almost seems like writing questions - in particular, anticipating all possible outcomes and then phrasing questions so there will be unambiguous resolutions given any possible outcome - is harder than assigning probabilities once you've already written the question.
I argue that 15% is on the lower end. The relevant base rate is near 20%, and four years is a long time in politics. I'd say it is too long to make strong predictions about specifics of he next presidential elections to deviate from the base rate more than 5 points.
Partisan tension may be high, but I'd argue that heightened tensions and general weirdness of recent times contribute towards general uncertainty about electoral results. In addition to established third parties, there is a non-zero chance for large scale party realignment or splits, and I can see substantial part of uncertainty splitting in that direction. Was there much reason in 1908 to suspect the great upset of 1912? There had been only one 5+% third party election in previous ~30 years, there was little indication that TR would run *against* Taft or that Debs would double his 1908/1904 level turnout. As a case of upset of smaller magnitude, Perot launched his 1992 campaign only after the primary season had started.
I would pick the general base rate as the best estimate that incorporates how often this kind of things may happen in US politics. And anyway, it still gives 80% chance for all 3rd candidates getting <5% of votes, which is the usual trend. 80% is often.
"9. Will China’s reach 100,000 official cases?
China's official case count on 12/31/20 was 95,963, so false."
Actually, depends. China publishes two different offical numbers. What you took is the number of symptomatic cases (which China, other than all other countries, enters into the official statistics websites). China also publishes the number of asymptomatic cases each day (positive test but no symptoms), but doesn't make them available in a nice and accessible format. If you count those, too, you are way above 100,000.
https://en.wikipedia.org/wiki/Statistics_of_the_COVID-19_pandemic_in_mainland_China
So the answer is both "true" and "false". Very much in the spirit of Yin and Yang.
Long thoughtful article by Siddhartha Mukherjee. "While the virus has ravaged rich nations, reported death rates in poorer ones remain relatively low. What probing this epidemiological mystery can tell us about global health."
https://www.newyorker.com/magazine/2021/03/01/why-does-the-pandemic-seem-to-be-hitting-some-countries-harder-than-others
I made my own predictions on Scott's questions. Assuming I've scored correctly with ln(p) - ln(.5), I managed a *3.34*!
1. Bay Area lockdown (eg restaurants closed) will be extended beyond June 15 20% -0.92
2. …until Election Day 10% -1.61
3. Fewer than 100,000 US coronavirus deaths 10% 1.61
4. Fewer than 300,000 US coronavirus deaths 70% -0.34
5. Fewer than 3 million US coronavirus deaths 90% 0.59
6. US has highest official death toll of any country 90% 0.59
7. US has highest death toll as per expert guesses of real numbers 80% 0.47
8. NYC widely considered worst-hit US city 80% 0.47
9. China’s (official) case number goes from its current 82,000 to 100,000 by the end of the year 70% -0.34
10. A coronavirus vaccine has been approved for general use and given to at least 10,000 people somewhere in the First World 20% -0.92
11. Best scientific consensus ends up being that hydroxychloroquine was significantly effective 20% 0.92
12. I personally will get coronavirus (as per my best guess if I had it; positive test not needed) 10% 1.61
13. Someone I am close to (housemate or close family member) will get coronavirus 30% 0.51
14. General consensus is that we (April 2020 US) were overreacting 10% 1.61
15. General consensus is that we (April 2020 US) were underreacting 30% 0.51
16. General consensus is that summer made coronavirus significantly less dangerous 10% -1.61
17. …and there is a catastrophic (50K+ US deaths, or more major lockdowns, after 1+ month without these things) 2nd wave in fall 30% -0.51
19. At least half of states send every voter a mail-in ballot in 2020 presidential election 10% 1.61
20. PredictIt is uncertain (less than 95%) who won the presidential election for more than 24 hours after Election Day 20% -0.92
"I will never bet against Zvi on anything. Last year he bet against me on what restaurant I would have dinner at, without knowing anything about my situation or food preferences, and won anyway."
I am so curious about this.