Looking at the first graph of the PredictIt market, 4% for the decision in Dobbs still seems awfully high. That's not based on any actual knowledge of the case or the court. It's based on the fact that "MI" stands for Michigan, not Mississippi.
Missed a main point: $14000 can turn a completely tertiary question into the biggest market. Or in other words, these prediction markets are eminently manipulable for minimal cost. This is a recipe for f***ery.
I never participated in the "what do you most disagree with Scott about?" thread, and I agree with your points regarding why prediction markets seem useful. But, I think if there is something I do often disagree about, it's the usefulness of measures for turning prediction markets into *drivers* of policy. I think there are some cases where this is probably a good idea, but I think that in a lot of cases, attempting to make the step from using them as an analytical tool to using them as a policymaking tool runs headlong into Goodhart's Law.
Why don't these markets post odds in the traditional American odds format? Seems like either a silly oversight, or maybe it's just needless exclusionary. If these sites are interested in getting more traffic and action, then a first step would be post odds in a format that the typical gambler can understand.; i.e., if something has a 60% of happening then it's -150, or if it has a 20% chance of happening it's +400.
>One weird coda to this story: the process of choosing the new administrator continues, and (surprisingly!) two of the leading candidates are female. Someone might be about to make a lot of money.
My worry about prediction markets has always been, what if that person is the person that decides who the next administrator will be?
That's the type of manipulation I'm much more worried about, here.
"the US government constantly disagrees with itself".
State governments, too. NY has a can deposit deposit of $0.05 so people don't just throw their cans on the side of the road. NY also has a prohibition on open alcoholic cans in an auto so people just throw their cans on the side of the road. It's not that the left hand doesn't know what the right hand is doing -- it's that the left hand doesn't *care* what the right hand is doing.
>One weird coda to this story: the process of choosing the new administrator continues, and (surprisingly!) two of the leading candidates are female. Someone might be about to make a lot of money.
Is insider trading a risk with prediction markets (i.e, someone involved in picking the admin is making a huge bet)? On the one hand I would think not, since if someone moves the market to what they know will actually happen that improves the market's accuracy. On the other hand, every bet has two sides, and people will stop playing in the market if they know you can't win without inside knowledge of what will happen.
Somewhat off topic, but it's actually kind of funny that you're just finding out today that cloned dogs are available because I actually just found that out today too. Actually, with the recent Kentucky Derby, it got me wondering whether cloned race horses are a thing, and it turns out that they actually are- and may be priced pretty competitively with traditionally-bred race horses.
1. I agree that sufficiently large prediction markets should be pretty accurate (at least on a certain type of question) and have the trust, aggregation, and clarity benefits you extol here. But we don't actually have any prediction markets like that. It's fine to play with the existing ones, and try to support their growth, but I think it's dangerous to index too hard on the specific predictions they make today.
2. In the case where some very rich person dumps a bunch of money into a large prediction market to manipulate it, and then a bunch of other people dump (in total) a similar amount of money into it to bring it back to a more accurate number, it seems like "a significant fraction of the economy is tied up in this one prediction" could cause problems, even though the market itself will have gone back to "normal".
One interesting thing about the ACX moderator market is that that's a market where someone can have private information that's *much* better than the public. So I'm wondering whether it was actually wrong to bet so heavily against the manipulator, given that for all the public knows they could have been the person who can choose the replacement.
Could Musk-owned Twitter host The Prediction Market That Actually Gains Traction - it will have the network, the high profile, the Musk genius. Makes sense to me. It could resolve the farce of Disinformation/Misinformation.
I wonder what an "Opinion and Debate" prediction market might look like. For instance, let's say you had a market in which the question is "Did Donald Trump inspire the Jan 6 attack on the Capitol?"
It's a matter of opinion not fact, but you could still have a betting market around it. The market could be a site in which public debate takes place over the question. The question could have a length of 3 months, after which, contracts close and payout and their current price. So such a market could close out at 53%, say, and that would be the payout.
Such markets would be open to manipulation, but so what? Would they be interesting markets? Would anyone participate in them? Would they mean anything at the end of the day?
"The future of abortion" sure is a little light on what overruling Roe v. Wade means for people and why we care. Would especially love an analysis on how federal abortion allowance would fit into the Archipelago and Atomic Communitarianism worldview.
re: misinformation, Prediction markets solve the problem of predicting what some authority will say at some future date, but not which authority to use for the settlement source. So it only works if the chosen authority eventually converges on the truth and actually reports it. Big media can't usually get away with lying about simple binary questions of fact, but if there's any ambiguity they can round it to whichever side they feel like and e.g. denounce the lab leak theory prematurely.
Funny enough, this project got started because of a short conversation I had with one of the founders (Austin) in the comments of a Mantic Monday post!
Scott Adams predicted specifically that Trump would win in 2016 *in a landslide*. He didn't outperform the prediction markets: his prediction was wrong.
The problem with prediction markets is that the best people won’t focus on events too far in the future. My friends are trading crypto futures. If they can’t make at least 30% pa on good enough volume, they won’t care.
Re. the theoretical Trump manipulation: skewing the prediction market with money might not work, because that can be viewed as an unexpected, unexplained deviation. But manufactured news in coordination with targeted media pieces, bold claims _and_ skewing the market with money might kind of work. Even more so, if a lot of people are watching the prediction markets as a gauge and trusting them, like the modern stock market.
Also, another problem which surely has been raised, but I'll do it again just in case. Insider trading, e.g. Scott buying tons of the "female moderator" option, and then appointing a female. I don't really see the solution to that. It can be outlawed, but will be extremely hard to police. I'm pretty sure this must have been discussed, so if someone could link me that, I'd be very grateful.
Doesn't the prediction of 10% chance of a female moderator strikes anybody else as just... wrong? I mean, if the user base is 90-10, making that prediction essentially boils down to believing that the next moderator will be chosen at random. That cannot possibly be true... right?
The fact that there are strong female candidates in contention (although we don't know among how many, I'd assume that it is less than 20) means that the probability was necessarily higher than 10%, and the secret better was more accurate than the market -- although I concede that 100% was obviously wrong as well.
Which means that the market going back down to 7% (lower than 10%!) is essentially saying that this prediction market, the one with the highest volume on the platform, is the least accurate (against the commonly held assumption that a higher volume will tend to higher precision). This seems to me the perfect case against prediction markets. Am I missing something obvious?
"any more than the stock market is competing with investment banks and market analysts."
I thought they were competing-- that money flows to the more profitable markets. This is complicated by people wanting different things like risk vs. security or getting things wrong, or working on different time scales.
Do we need a theory for why there are different sorts of money markets?
Elon Musk made a fuss recently about index funds. Saying: if most of the market is index funds instead of people actually trading, then the market price records a lot of noise instead of an actual valuation of the various companies.
There's the same effect for prediction market vs experts. If you have a domain where the correct answer is found in the mouth of experts, then the prediction market is just the opinion of the expert plus noise. And the more people listen do markets instead of directly going to experts, the more noise I'm expecting to see in the markets.
It doesn't strike me as particularly impressive that a prediction market can copy the performance of an expert? I mean, we already have the expert? The market adds nothing in this case.
I think it's likely that prediction markets are overrated, at least until you add insider information, but then that makes for a whole different problem instead.
As someone who has been betting on political markets on BetFair Exchange (effectively a prediction market) for nearly 20 years, I'll say a couple of things about market manipulation.
First, someone spending a lot of money on a market to manipulate the price can do so, but they need to keep pumping in money to keep the price up, and the market will attract more and more "smart" money so it gets increasingly expensive to do so. This was done, intentionally, by one of the campaigns for the leader of the Liberal Democrats in 2006, and estimates were that they spent about half a million (sterling), and ran out of money about two weeks before votes were cast. This did, however, probably elevate them as the main challenger, though they still lost - but they would likely have finished third without this.
Second, the weight of each person's opinion depends on how much money they have; the 2020
US Presidential election market was not resolved until the electoral college met and voted (in December). Large numbers of Trump supporters kept the Trump probability around 10% from when the TV networks called the election until after. Like many people, I put in a few thousand to extract an easy 10% ROI, but the Trump supporters were putting in tens of millions and there just wasn't enough "smart" money to push the probabilities any lower.
It is possible that a really large amount of money can manipulate the market, either by a single person or by enough people that have been fooled by misinformation or disinformation. But it is rare and not having a restriction on insider trading (there's no legal restriction) means that the markets are probably more accurate. I know of many cases of political insiders making extremely healthy profits by betting on insider information about likely election results. Campaign staffers often use this to make up for relatively poor salaries.
My issue with prediction markets is that the arguments for them often don't distinguish between theoretical prediction markets with perfect liquidity and an infinite number of participants etc vs the prediction markets that are likely to exist in the real world. You can maybe get close to that with markets on extremely major issues like presidential elections, but that doesn't necessarily transfer to everything else. And proving how it would transfer tends to be elided
Scott, I don't mean to accuse you of dishonesty, just a normal oversight, but the last line of the Discorcasting section seems very close to a "heads I win, tails you lose" situation. Are you saying now, before it's decided, that if a female moderator is selected it would be strong evidence *against* the accuracy of prediction markets? If not, I don't think we can take the low price as evidence for prediction markets' resistance to manipulation.
Question about prediction markets generally: is there a good explanation for how/why they will handle long-term, unlikely events?
For instance, PredictIt offers shares that Bernie Sanders will *not* be the 2024 dem presidential nominee for $0.97. Payout is $1.00, so ~3% increase. Let's say I don't think Bernie has a shot -- I could buy a ton of shares, but I'm effectively locking my money up for *two years* for a 3% return. That's barely better than treasury bond yields over 2 years, and there are probably better places to park my money.
For longer-term predictions with even longer shots, this problem seems even bigger. It seems like prediction markets will necessarily be pretty off for these kinds of predictions, where they cap out at maybe 20:1 odds. Does that sound right? Are there other factors I'm not thinking about that will keep them accurate for this?
> Disclaimer that Manifold probably can’t handle probabilities this small correctly and there’s no reason to think 0.2% is more realistic than 2%. It’s not 10% though.
Is there any reason to think 0.2% is more realistic than 0.002% ?
I don't understand why you don't cover sports betting, Scott. Sports books are mature prediction markets with lots and lots of money sloshing around in them. Are they good? I dunno, I haven't looked into it, I'm in the camp that all this is kinda dumb. But it seems like a very good arena in which to test your claims for prediction markets. For instance, "public teams" (teams that have chronically inaccurate lines because people like to bet on them independent of their quality; the Yankees, Packers, and whatever team LeBron is on at the moment are public teams) are a notorious facet of sports betting that seem pretty bad for your claims about prediction markets. But I dunno, maybe public teams are a myth, or they were a thing in the 90s that got driven out by the rise of quantitatively sophisticated professional gamblers.
I know you find sports gauche, but if you don't care enough about prediction markets to overcome that, I don't see why I should be asked to care about them either.
Prediction markets kind of remind me of usury. Usury, of course, being the practice of lending money at interest, which was widely regarded in the western world as a sin for over a thousand years. The sin, very generally speaking, was loan sharks: someone loans you money at interest, you can't pay it back, you're in debt to them forever. It was also seen as cheating people: you (the lender) don't actually do any work, so why should you get paid? Today that objection sounds silly: they get paid to compensate the risk of not being paid back, and loans provide a lot of social good so what's wrong with making a profit? But back then all they had as their model of loaning money at interest was loan sharks, so it was condemned as a sinful practice. Today the world economy runs off of loaning money at interest, and the idea of banning it would be almost unthinkable.
So to with prediction markets. Gambling is considered a sin, and for good reason. We all know people who have ruined themselves through gambling, and it's hard to see what social benefit gambling provides anyone besides entertainment. It makes sense that just about every religion and most respectable secular authorities condemn gambling as evil. Yet I can imagine a future, a few hundred years from now, where the idea that people once considered betting on the outcome of events as a terrible sin will be just as surprising as when someone today finds out that lending money at interest was once considered a terrible sin. It's not that our values will change, but the form of the thing itself will change. Just as the free market transforms avarice into societal wealth generation, prediction markets could transform gambling into a prosocial good as well.
> A lot of times, people ask: what if some very rich person tried to manipulate a prediction market? Like, what if there was a market on whether Trump would win the presidency, and Trump himself dropped $100 million into making it say YES, either out of vanity or in the hopes that people would think he was “inevitable” and stop resisting.
> And the theory has always been: if traders see a clearly mispriced market, they’ll rush to it and correct the mispricing in order to make easy money. Trump would be out $100 million and the market would stay correct.
> We’ve never been able to test the theory, and until someone drops $100 million into a real money market, we never will.
Another edition of "please someone explain betting markets to me." In this edition, I made a bet on Manifold and am super confused by the result.
I think there's a solid chance (maybe 20%?) that Republicans will make Abortion illegal nationally within the next 5 years. I saw that the Manifold market was at 6% with $198 bet, so I decided to weigh in. I submitted $25 for a Yes bet.
Two confusing things happened: First, I received 111 shares of "Yes." I would expect 416 shares ($25 divided by $0.06 per share). Why did the shares cost $0.23 instead of $0.06?
Secondly, the Manifold probability jumped from 6% to 78%! That's a huge jump! One, it's above my own true estimate, which is dumb. Two, my $25 bet is about 1/9th of the $223 total market value, so why is my Yes bet outweighing everyone else? Looking at the graph, all the other bets (totaling $198) moved the probability up or down by a couple percent. What's going on?
Did anyone sign up for the forecasting tournament mentioned in the previous open thread? If so, did you get a confirmation email after submitting the signup form?
This may be a pointless bit of language carping, but I wonder if it's worth considering a slight change in language, away from terms like wrong and right, when you make probabilistic predictions.
"By the way, in 2018, I got this horribly wrong - I said only a 1% chance of Roe v. Wade getting repealed in the next 5 years."
It's not obvious to me that this prediction was wrong! I mean, I understand this use of the term, but strictly speaking, the claim that you made has not been invalidated or disproven by subsequent events. And part of the value of prediction markets is precisely to push people away from simplistic right/wrong pigeonholing of views. One way to do that is to not use the simplistic language when discussing them. Instead you could use something like, "This now looks like a poor prediction; I lost that bet; in hindsight, it looks like I may have evaluated some factors incorrectly." The potential downside is that you sound a bit less human/normal, of course, so it's bit of a trade off.
> On the DSL threat, someone brought up that the US government was quicker to call the Russian invasion of Ukraine than most prediction markets. I agree this happened. In some sense, it’s unsurprising; the US government has spy satellites, moles in the Kremlin, and lots of highly-paid analysts. Of course they should do better than everyone else.
Given that I seem to have a pretty good track record of prediction, it's a shame that I've been banned from both Astral Codex Ten as well as the Motte. The consistent pattern there seems to be small-minded mods telling me that I'm not NICE enough, but as we learned from the Ukraine invasion, reality is not always NICE. My embrace of reality in preference to NICENESS could have averted a war and a pandemic, but because people don't like taking advice from somebody who isn't NICE, I guess we needed to have a whole bunch of people die instead. Maybe the community should stop optimizing for niceness? It doesn't seem to be a very useful (or rational) strategy.
Get a majority of administrator candidates on board. Buy. Candidates announce they now identify as female. Sell. Candidates announce they now identify as male. Buy...
(Admittedly this works better if the market is real money and you can offer them a percentage of the profits. But I personally would try to get in on the first gender market manipulation scheme just for the anecdote.)
I have not read all the comments / posts / etc related to prediction markets, so apologies if this is redundant/covered. But I have not seen it covered in what I have read.
Prediction markets seem neat now, because they seem to be high-signal. People mostly seem to want to predict, not manipulate. People mostly want to be accurate according to factual reality.
This seems to mirror the early Internet. Very less overall content but higher average quality. But then came mass adoption, Facebook, Twitter. More recently, concerted, sustained disinformation.
I would be concerned what happens if prediction markets become less niche, more noticed: "successful" by some metrics. Say Russia or the CCP or some other government-level entity with enormous resources takes an interest in manipulation. The goal here might not be to drive the outcome, merely to muddy the waters and maximize noise. So you could not simply get rich by disagreeing.
on Xi re-elected CCP General Secretary: as somebody who bought No on Polymarket, I see the main possibility to be him retaining power but with a different title. Think of Putin's 'castling' with Medvedev as a comparison. It's not probable, but IMO it's over 10%
Aubrey de Grey says that there are seven (and only seven) things that make us age, and he has solutions (for some understanding of "solution") for all seven.
So
(a) do you trust Aubrey de Grey? Do you trust him more or less after seeing a picture of him? Opinions, as they say, vary.
(b) do you trust that his solutions will actually work (I'm assuming the next hundred years or so are long enough to iron out any engineering issues; the conceptual issues are the point)
(c) [and this is the big one] do you believe that "society" will allow these solutions to be tested then implemented? I flip back and forth on this. On the one hand, rich people and other countries are going to be happy to let this go where it goes. On the other hand, "society" has a long history of being fairly successful in slowing down or suppressing certain types of medical research, especially stuff that requires large scale effort not just a mad scientist in lab (eg much interventionist genetic medicine). So...
Looking at the first graph of the PredictIt market, 4% for the decision in Dobbs still seems awfully high. That's not based on any actual knowledge of the case or the court. It's based on the fact that "MI" stands for Michigan, not Mississippi.
Missed a main point: $14000 can turn a completely tertiary question into the biggest market. Or in other words, these prediction markets are eminently manipulable for minimal cost. This is a recipe for f***ery.
I never participated in the "what do you most disagree with Scott about?" thread, and I agree with your points regarding why prediction markets seem useful. But, I think if there is something I do often disagree about, it's the usefulness of measures for turning prediction markets into *drivers* of policy. I think there are some cases where this is probably a good idea, but I think that in a lot of cases, attempting to make the step from using them as an analytical tool to using them as a policymaking tool runs headlong into Goodhart's Law.
"once they became important enough to matter, biased interests would take them over"
Once [prediction markets] became important enough to matter, [the US government] would take them over.
Why don't these markets post odds in the traditional American odds format? Seems like either a silly oversight, or maybe it's just needless exclusionary. If these sites are interested in getting more traffic and action, then a first step would be post odds in a format that the typical gambler can understand.; i.e., if something has a 60% of happening then it's -150, or if it has a 20% chance of happening it's +400.
Doesn't Rich Strike winning the Kentucky Derby prove prediction markets don't work??
TIL of the ACX discord server. Anybody have an invite link, please?
Typo thread:
I think “predicting the results of the recent Supreme Court *link*.” was meant to be “predicting the results of the recent Supreme Court *leak*.”
>One weird coda to this story: the process of choosing the new administrator continues, and (surprisingly!) two of the leading candidates are female. Someone might be about to make a lot of money.
My worry about prediction markets has always been, what if that person is the person that decides who the next administrator will be?
That's the type of manipulation I'm much more worried about, here.
"the US government constantly disagrees with itself".
State governments, too. NY has a can deposit deposit of $0.05 so people don't just throw their cans on the side of the road. NY also has a prohibition on open alcoholic cans in an auto so people just throw their cans on the side of the road. It's not that the left hand doesn't know what the right hand is doing -- it's that the left hand doesn't *care* what the right hand is doing.
>One weird coda to this story: the process of choosing the new administrator continues, and (surprisingly!) two of the leading candidates are female. Someone might be about to make a lot of money.
Is insider trading a risk with prediction markets (i.e, someone involved in picking the admin is making a huge bet)? On the one hand I would think not, since if someone moves the market to what they know will actually happen that improves the market's accuracy. On the other hand, every bet has two sides, and people will stop playing in the market if they know you can't win without inside knowledge of what will happen.
> One weird coda to this story: the process of choosing the new administrator continues, and (surprisingly!) two of the leading candidates are female.
Two out of how many?
Somewhat off topic, but it's actually kind of funny that you're just finding out today that cloned dogs are available because I actually just found that out today too. Actually, with the recent Kentucky Derby, it got me wondering whether cloned race horses are a thing, and it turns out that they actually are- and may be priced pretty competitively with traditionally-bred race horses.
Here's the article I read (though I cannot vouch for the trustworthiness of the publication), please remove if links aren't allowed: https://robbreport.com/lifestyle/sports-leisure/cloned-horses-competition-2881549/
1. I agree that sufficiently large prediction markets should be pretty accurate (at least on a certain type of question) and have the trust, aggregation, and clarity benefits you extol here. But we don't actually have any prediction markets like that. It's fine to play with the existing ones, and try to support their growth, but I think it's dangerous to index too hard on the specific predictions they make today.
2. In the case where some very rich person dumps a bunch of money into a large prediction market to manipulate it, and then a bunch of other people dump (in total) a similar amount of money into it to bring it back to a more accurate number, it seems like "a significant fraction of the economy is tied up in this one prediction" could cause problems, even though the market itself will have gone back to "normal".
One interesting thing about the ACX moderator market is that that's a market where someone can have private information that's *much* better than the public. So I'm wondering whether it was actually wrong to bet so heavily against the manipulator, given that for all the public knows they could have been the person who can choose the replacement.
Scott, I'd be interested to hear any thoughts you might have about the ideas in this post: https://www.lesswrong.com/posts/GNJNJ2cRyvSLp2owH/speculations-concerning-the-first-free-ish-prediction-market
It's a year old, but I stand by it (aside from the Kalshi optimism, which turned out to be misplaced).
Could Musk-owned Twitter host The Prediction Market That Actually Gains Traction - it will have the network, the high profile, the Musk genius. Makes sense to me. It could resolve the farce of Disinformation/Misinformation.
I wonder what an "Opinion and Debate" prediction market might look like. For instance, let's say you had a market in which the question is "Did Donald Trump inspire the Jan 6 attack on the Capitol?"
It's a matter of opinion not fact, but you could still have a betting market around it. The market could be a site in which public debate takes place over the question. The question could have a length of 3 months, after which, contracts close and payout and their current price. So such a market could close out at 53%, say, and that would be the payout.
Such markets would be open to manipulation, but so what? Would they be interesting markets? Would anyone participate in them? Would they mean anything at the end of the day?
"The future of abortion" sure is a little light on what overruling Roe v. Wade means for people and why we care. Would especially love an analysis on how federal abortion allowance would fit into the Archipelago and Atomic Communitarianism worldview.
re: misinformation, Prediction markets solve the problem of predicting what some authority will say at some future date, but not which authority to use for the settlement source. So it only works if the chosen authority eventually converges on the truth and actually reports it. Big media can't usually get away with lying about simple binary questions of fact, but if there's any ambiguity they can round it to whichever side they feel like and e.g. denounce the lab leak theory prematurely.
(A bit of self-promotion, hope that's okay)
Manifold is now allowing users to send money to charity:
https://manifold.markets/charity
Funny enough, this project got started because of a short conversation I had with one of the founders (Austin) in the comments of a Mantic Monday post!
Scott Adams predicted specifically that Trump would win in 2016 *in a landslide*. He didn't outperform the prediction markets: his prediction was wrong.
The problem with prediction markets is that the best people won’t focus on events too far in the future. My friends are trading crypto futures. If they can’t make at least 30% pa on good enough volume, they won’t care.
Re. the theoretical Trump manipulation: skewing the prediction market with money might not work, because that can be viewed as an unexpected, unexplained deviation. But manufactured news in coordination with targeted media pieces, bold claims _and_ skewing the market with money might kind of work. Even more so, if a lot of people are watching the prediction markets as a gauge and trusting them, like the modern stock market.
Also, another problem which surely has been raised, but I'll do it again just in case. Insider trading, e.g. Scott buying tons of the "female moderator" option, and then appointing a female. I don't really see the solution to that. It can be outlawed, but will be extremely hard to police. I'm pretty sure this must have been discussed, so if someone could link me that, I'd be very grateful.
Doesn't the prediction of 10% chance of a female moderator strikes anybody else as just... wrong? I mean, if the user base is 90-10, making that prediction essentially boils down to believing that the next moderator will be chosen at random. That cannot possibly be true... right?
The fact that there are strong female candidates in contention (although we don't know among how many, I'd assume that it is less than 20) means that the probability was necessarily higher than 10%, and the secret better was more accurate than the market -- although I concede that 100% was obviously wrong as well.
Which means that the market going back down to 7% (lower than 10%!) is essentially saying that this prediction market, the one with the highest volume on the platform, is the least accurate (against the commonly held assumption that a higher volume will tend to higher precision). This seems to me the perfect case against prediction markets. Am I missing something obvious?
"any more than the stock market is competing with investment banks and market analysts."
I thought they were competing-- that money flows to the more profitable markets. This is complicated by people wanting different things like risk vs. security or getting things wrong, or working on different time scales.
Do we need a theory for why there are different sorts of money markets?
Elon Musk made a fuss recently about index funds. Saying: if most of the market is index funds instead of people actually trading, then the market price records a lot of noise instead of an actual valuation of the various companies.
There's the same effect for prediction market vs experts. If you have a domain where the correct answer is found in the mouth of experts, then the prediction market is just the opinion of the expert plus noise. And the more people listen do markets instead of directly going to experts, the more noise I'm expecting to see in the markets.
So what’s to stop the outgoing discord admin from betting on his own choice?
It doesn't strike me as particularly impressive that a prediction market can copy the performance of an expert? I mean, we already have the expert? The market adds nothing in this case.
I think it's likely that prediction markets are overrated, at least until you add insider information, but then that makes for a whole different problem instead.
As someone who has been betting on political markets on BetFair Exchange (effectively a prediction market) for nearly 20 years, I'll say a couple of things about market manipulation.
First, someone spending a lot of money on a market to manipulate the price can do so, but they need to keep pumping in money to keep the price up, and the market will attract more and more "smart" money so it gets increasingly expensive to do so. This was done, intentionally, by one of the campaigns for the leader of the Liberal Democrats in 2006, and estimates were that they spent about half a million (sterling), and ran out of money about two weeks before votes were cast. This did, however, probably elevate them as the main challenger, though they still lost - but they would likely have finished third without this.
Second, the weight of each person's opinion depends on how much money they have; the 2020
US Presidential election market was not resolved until the electoral college met and voted (in December). Large numbers of Trump supporters kept the Trump probability around 10% from when the TV networks called the election until after. Like many people, I put in a few thousand to extract an easy 10% ROI, but the Trump supporters were putting in tens of millions and there just wasn't enough "smart" money to push the probabilities any lower.
It is possible that a really large amount of money can manipulate the market, either by a single person or by enough people that have been fooled by misinformation or disinformation. But it is rare and not having a restriction on insider trading (there's no legal restriction) means that the markets are probably more accurate. I know of many cases of political insiders making extremely healthy profits by betting on insider information about likely election results. Campaign staffers often use this to make up for relatively poor salaries.
My issue with prediction markets is that the arguments for them often don't distinguish between theoretical prediction markets with perfect liquidity and an infinite number of participants etc vs the prediction markets that are likely to exist in the real world. You can maybe get close to that with markets on extremely major issues like presidential elections, but that doesn't necessarily transfer to everything else. And proving how it would transfer tends to be elided
What about prediction of Moldova/Romania and Ukraine cooperating and ending Transnistria issue with force?
Scott, I don't mean to accuse you of dishonesty, just a normal oversight, but the last line of the Discorcasting section seems very close to a "heads I win, tails you lose" situation. Are you saying now, before it's decided, that if a female moderator is selected it would be strong evidence *against* the accuracy of prediction markets? If not, I don't think we can take the low price as evidence for prediction markets' resistance to manipulation.
Question about prediction markets generally: is there a good explanation for how/why they will handle long-term, unlikely events?
For instance, PredictIt offers shares that Bernie Sanders will *not* be the 2024 dem presidential nominee for $0.97. Payout is $1.00, so ~3% increase. Let's say I don't think Bernie has a shot -- I could buy a ton of shares, but I'm effectively locking my money up for *two years* for a 3% return. That's barely better than treasury bond yields over 2 years, and there are probably better places to park my money.
For longer-term predictions with even longer shots, this problem seems even bigger. It seems like prediction markets will necessarily be pretty off for these kinds of predictions, where they cap out at maybe 20:1 odds. Does that sound right? Are there other factors I'm not thinking about that will keep them accurate for this?
> Disclaimer that Manifold probably can’t handle probabilities this small correctly and there’s no reason to think 0.2% is more realistic than 2%. It’s not 10% though.
Is there any reason to think 0.2% is more realistic than 0.002% ?
I don't understand why you don't cover sports betting, Scott. Sports books are mature prediction markets with lots and lots of money sloshing around in them. Are they good? I dunno, I haven't looked into it, I'm in the camp that all this is kinda dumb. But it seems like a very good arena in which to test your claims for prediction markets. For instance, "public teams" (teams that have chronically inaccurate lines because people like to bet on them independent of their quality; the Yankees, Packers, and whatever team LeBron is on at the moment are public teams) are a notorious facet of sports betting that seem pretty bad for your claims about prediction markets. But I dunno, maybe public teams are a myth, or they were a thing in the 90s that got driven out by the rise of quantitatively sophisticated professional gamblers.
I know you find sports gauche, but if you don't care enough about prediction markets to overcome that, I don't see why I should be asked to care about them either.
Prediction markets kind of remind me of usury. Usury, of course, being the practice of lending money at interest, which was widely regarded in the western world as a sin for over a thousand years. The sin, very generally speaking, was loan sharks: someone loans you money at interest, you can't pay it back, you're in debt to them forever. It was also seen as cheating people: you (the lender) don't actually do any work, so why should you get paid? Today that objection sounds silly: they get paid to compensate the risk of not being paid back, and loans provide a lot of social good so what's wrong with making a profit? But back then all they had as their model of loaning money at interest was loan sharks, so it was condemned as a sinful practice. Today the world economy runs off of loaning money at interest, and the idea of banning it would be almost unthinkable.
So to with prediction markets. Gambling is considered a sin, and for good reason. We all know people who have ruined themselves through gambling, and it's hard to see what social benefit gambling provides anyone besides entertainment. It makes sense that just about every religion and most respectable secular authorities condemn gambling as evil. Yet I can imagine a future, a few hundred years from now, where the idea that people once considered betting on the outcome of events as a terrible sin will be just as surprising as when someone today finds out that lending money at interest was once considered a terrible sin. It's not that our values will change, but the form of the thing itself will change. Just as the free market transforms avarice into societal wealth generation, prediction markets could transform gambling into a prosocial good as well.
> A lot of times, people ask: what if some very rich person tried to manipulate a prediction market? Like, what if there was a market on whether Trump would win the presidency, and Trump himself dropped $100 million into making it say YES, either out of vanity or in the hopes that people would think he was “inevitable” and stop resisting.
> And the theory has always been: if traders see a clearly mispriced market, they’ll rush to it and correct the mispricing in order to make easy money. Trump would be out $100 million and the market would stay correct.
> We’ve never been able to test the theory, and until someone drops $100 million into a real money market, we never will.
The problem is that you won't necessarily know whether this has been done. It's pretty plausible that Michael Bloomberg did exactly what you said Trump could have done (see https://www.vox.com/future-perfect/2020/2/14/21137882/prediction-markets-bloomberg-sanders-president), but we don't know.
A lot of commenters are making fun of this [someone born before 2001 living to 150] for being too low.
Increases in life expectancy have stalled or reversed recently.
Another edition of "please someone explain betting markets to me." In this edition, I made a bet on Manifold and am super confused by the result.
I think there's a solid chance (maybe 20%?) that Republicans will make Abortion illegal nationally within the next 5 years. I saw that the Manifold market was at 6% with $198 bet, so I decided to weigh in. I submitted $25 for a Yes bet.
Two confusing things happened: First, I received 111 shares of "Yes." I would expect 416 shares ($25 divided by $0.06 per share). Why did the shares cost $0.23 instead of $0.06?
Secondly, the Manifold probability jumped from 6% to 78%! That's a huge jump! One, it's above my own true estimate, which is dumb. Two, my $25 bet is about 1/9th of the $223 total market value, so why is my Yes bet outweighing everyone else? Looking at the graph, all the other bets (totaling $198) moved the probability up or down by a couple percent. What's going on?
Did anyone sign up for the forecasting tournament mentioned in the previous open thread? If so, did you get a confirmation email after submitting the signup form?
>Manifold recently ran an unintentional, kind of crazy experiment in manipulating prediction markets.
Alternative possibility, someone knows who the moderator is going to be and wants to have a massive position on that outcome.
This may be a pointless bit of language carping, but I wonder if it's worth considering a slight change in language, away from terms like wrong and right, when you make probabilistic predictions.
"By the way, in 2018, I got this horribly wrong - I said only a 1% chance of Roe v. Wade getting repealed in the next 5 years."
It's not obvious to me that this prediction was wrong! I mean, I understand this use of the term, but strictly speaking, the claim that you made has not been invalidated or disproven by subsequent events. And part of the value of prediction markets is precisely to push people away from simplistic right/wrong pigeonholing of views. One way to do that is to not use the simplistic language when discussing them. Instead you could use something like, "This now looks like a poor prediction; I lost that bet; in hindsight, it looks like I may have evaluated some factors incorrectly." The potential downside is that you sound a bit less human/normal, of course, so it's bit of a trade off.
> On the DSL threat, someone brought up that the US government was quicker to call the Russian invasion of Ukraine than most prediction markets. I agree this happened. In some sense, it’s unsurprising; the US government has spy satellites, moles in the Kremlin, and lots of highly-paid analysts. Of course they should do better than everyone else.
I predicted this here:
https://www.reddit.com/r/TheMotte/comments/s2gk0v/will_nato_expansionism_lead_to_a_war_between_the/
I also hinted at Covid ahead of time also in my book:
https://www.amazon.com/Dark-Arts-Rationality-Updated-Digital-ebook/dp/B07ZGHS73F
Given that I seem to have a pretty good track record of prediction, it's a shame that I've been banned from both Astral Codex Ten as well as the Motte. The consistent pattern there seems to be small-minded mods telling me that I'm not NICE enough, but as we learned from the Ukraine invasion, reality is not always NICE. My embrace of reality in preference to NICENESS could have averted a war and a pandemic, but because people don't like taking advice from somebody who isn't NICE, I guess we needed to have a whole bunch of people die instead. Maybe the community should stop optimizing for niceness? It doesn't seem to be a very useful (or rational) strategy.
Get a majority of administrator candidates on board. Buy. Candidates announce they now identify as female. Sell. Candidates announce they now identify as male. Buy...
(Admittedly this works better if the market is real money and you can offer them a percentage of the profits. But I personally would try to get in on the first gender market manipulation scheme just for the anecdote.)
A core issue with prediction markets is that the smartest people in the world are not particularly motivated by money.
The court-packing Metaculus question has moved up since this post was written.
I have not read all the comments / posts / etc related to prediction markets, so apologies if this is redundant/covered. But I have not seen it covered in what I have read.
Prediction markets seem neat now, because they seem to be high-signal. People mostly seem to want to predict, not manipulate. People mostly want to be accurate according to factual reality.
This seems to mirror the early Internet. Very less overall content but higher average quality. But then came mass adoption, Facebook, Twitter. More recently, concerted, sustained disinformation.
I would be concerned what happens if prediction markets become less niche, more noticed: "successful" by some metrics. Say Russia or the CCP or some other government-level entity with enormous resources takes an interest in manipulation. The goal here might not be to drive the outcome, merely to muddy the waters and maximize noise. So you could not simply get rich by disagreeing.
on Xi re-elected CCP General Secretary: as somebody who bought No on Polymarket, I see the main possibility to be him retaining power but with a different title. Think of Putin's 'castling' with Medvedev as a comparison. It's not probable, but IMO it's over 10%
Re aging:
Aubrey de Grey says that there are seven (and only seven) things that make us age, and he has solutions (for some understanding of "solution") for all seven.
So
(a) do you trust Aubrey de Grey? Do you trust him more or less after seeing a picture of him? Opinions, as they say, vary.
(b) do you trust that his solutions will actually work (I'm assuming the next hundred years or so are long enough to iron out any engineering issues; the conceptual issues are the point)
(c) [and this is the big one] do you believe that "society" will allow these solutions to be tested then implemented? I flip back and forth on this. On the one hand, rich people and other countries are going to be happy to let this go where it goes. On the other hand, "society" has a long history of being fairly successful in slowing down or suppressing certain types of medical research, especially stuff that requires large scale effort not just a mad scientist in lab (eg much interventionist genetic medicine). So...