Poll: is betting on a yes/no market after the outcome is known bad?
9%
chance
Jun 25
M$465 bet
Do you think it's bad to bet (or allow betting) on a market after the outcome is known? For this question, please consider only binary (yes/no) markets. Vote by commenting with YES, NO, or a percentage at the start of your comment. (YES = 100%, NO = 0%, you can also choose any percentage in between if you think it's e.g. 30% bad.) The market resolves to the average of valid votes in the poll as of 1 day after market close. There are multiple perspectives you could consider, including the traders, author, liquidity providers, and readers. Different people may prioritize different things, so I'm asking whether you think it's bad overall. A little context for those not familiar with the market math: for binary (yes/no) markets, buying when the outcome is known lets you take some additional profits, which comes from the liquidity providers. But it does not affect any previous traders already holding shares, since if you make a trade and hold until resolution, any later trades do not affect your profits/losses. See related market with some discussion by Austin: https://manifold.markets/Austin/will-manifolds-fixedpayout-markets You can update your vote by commenting again; only your last response will be counted.
Predictor
Predictor is betting NO at 9%
NO
0
Predictor
Predictor bought M$10 of NO
NO
0
jack
Jack is betting NO at 9%
I also have another poll question: is your opinion any different about whether it's bad for the author to bet on their market just before resolving it? (Again, for fixed-payout yes/no markets only.) If there's sufficient interest in this I may make another market for it.
0
jack
Jack is betting NO at 9%
NO (mostly) Good: I think it's useful for the market to quickly aggregate the information about fact that the outcome has become known, as the author might not be around to resolve it right away. Betting rewards those who see the news and update the market to reflect it the fastest. And sometimes, it's hard to tell when the market should resolve, and the fact that you can earn a profit by figuring it out is good. And as I mentioned, it's not harmful to other traders in fixed-payout (CFMM) markets. Bad: It incentivizes people to keep information they have about the resolution private in order to take more profits. Example: https://manifold.markets/Phi/the-prime-factors-of-n-are-both-2-m. Multiple traders (including myself) factored the number, but we did not reveal this information for several days because the market incentivizes us not to! It's not clear how you could adjust the market mechanism to avoid this issue though. This is a general issue with prediction markets - you're incentivized to keep private information private to make more profits, which is counter to the goal of most efficiently aggregating information. Also, it provides large profit incentives for traders to spend more time on closely following events that are about to resolve (e.g. will stock X close above Y today, will X win the game, etc). And I think this is not a very useful use of the human capital on prediction markets, although it can be fun and engaging. And it's bad for the liquidity providers. If the outcome will be known at a specific time, they can withdraw liquidity in advance, but that doesn't work for many markets e.g. "when will event X happen". Liquidity providers selling YES basically will always lose whatever chunk of M$ they committed if event X happens and traders buy YES before they can withdraw the liquidity - which means LPs would be very conservative about selling YES shares, which means worse liquidity for traders. One possible option is to have a mechanism that retroactively closes the market just before event X happened, and N/A-s all trades after that point. (Metaculus has a mechanism like this.)
0
Kronopath
Kronopath bought M$50 of NO
NO. A big part of the benefit of prediction markets is incentivizing research and keeping up with news. If you make a market about whether the Queen of England was going to die this year, and the market swings to 99% all of a sudden because of a breaking news story, that’s useful information for anyone watching that market. What’s the alternative, humbly pass on the profits and wait for the market creator to discover the news?
0
FranekZak
NO. Unless you have private information, then maybe.
0
bcongdon
Ben is betting NO at 24%
@FranekZak Arguably even if you have private information, you're making that information public by betting on it, which is (in many cases) a public good
0
FranekZak
Franek Żak bought M$5 of NO
@bcongdon In many cases yes, although I’d argue that sometimes its not great.
0
ScottLawrence
Scott Lawrence is betting NO at 15%
@FranekZak Care to give a specific example of when it's not great (particularly in the context of the low stakes of fake internet points)? Is it just "I don't want people to be incentivized to keep information private", or is there something more that I'm missing?
0
jack
Jack is betting NO at 9%
@ScottLawrence Here's an example: https://manifold.markets/Phi/the-prime-factors-of-n-are-both-2-m. Multiple people (including myself) figured out the answer minutes after the market was created, but we kept mum in order to take more profits. This is a rational response to the market profit incentives and I don't think it's bad in the sense of being unethical, but it's clearly bad in the sense of the market working well. I think it highlights a general issue with the structure of prediction markets - the market incentivizes you to keep private information private in order to make more profits, which is counter to the goal of more efficiently aggregating information and predictions.
0
MichaelWheatley
@jack Is that so? Isn't the optimal strategy to bet it down to the correct probability, keeping your reasons hidden so you can keep buying it back down, and then once you run out of liquidity, revealing your knowledge so you can cash out? The key point being that it's staying at the correct probability at all times because you're either buying it back down there or revealing the truth?
0
jack
Jack is betting NO at 9%
@MichaelWheatley No, because a) you can make more money by predicting and trading on price movements, not just the final resolution, and b) other people look at the betting pattern and infer information from that, and so you can make more money by misleading/misdirecting them. This market is an example: Nuno was the trader repeatedly betting M1000 NO (which is basically the strategy of keeping your reasons hidden while betting down to the correct probability), and made a good profit off that, but I made even more profit by betting down to ~20-30% and buying YES when the probability dropped to <1% and then selling the YES when it went up. I could do this because I predicted that there were several betters who had a significant belief that the number was not easily factorizable and were betting towards 50%. And by only betting down a moderate amount, I better hid my private information, while by betting down to nearly 0 Nuno was effectively claiming that they knew the answer was NO.
0
ScottLawrence
Scott Lawrence bought M$30 of NO
NO! It's great! When creating markets, I usually try to set the close date somewhat after the result is known, to incentivize other people to help me find evidence about how it should resolve. Previous participants aren't harmed, I benefit, the people who make later bets benefit. The only drawback I'm aware of is that people who are just good at finding/acting on information quickly, rather than good at predicting, become M$-rich. I don't know quite what I think about that, but I'm rather against weighting "society"-wide considerations like that more strongly than "it benefits everyone involved".
0
bcongdon
Ben bought M$10 of NO
NO. Liquidity providers could set an arbitrarily early close time to avoid this, in many cases. Otherwise, the onus is on them to close the market when the outcome becomes known
0
bcongdon
Ben is betting NO at 9%
Also, Manifold’s Yes/No markets are fixed payout, which means that you’re not “stealing” profits from others by betting after the outcome is known. Free response markets don’t have this property, so I’d be more reluctant in that case
0