Followup is important; I notice that Eliezer has not exactly gone around trumpeting Wei Dai’s followup comment where he mentions losing almost all of his coronavirus profits as evidence that maybe the EMH is right after all
When Eliezer posted about my bet, I had only gained 7x my initial bet (and that’s what he posted about), and although I ended up losing 80% of my paper profits (which were 50x at one point) I still gained 10x my initial bet. So him not posting further followup seems fine? And at least from my personal perspective (i.e., where selection bias isn’t an issue) the final outcome still seems to be strong evidence against EMH.
the final outcome still seems to be strong evidence against EMH
It seems to me that the EMH is a spherical cow. In reality, there has to be some process by which market prices comes to reflect new information. And whoever figures it out first, either because they had private information, or because they did better analysis, should make some profit.
And the question might be—is it always the same people who figure it out first, or is there some randomness involved? (Consider a toy model where N people are trading one stock and every day a fairy randomly selects 1 of the N people and tells them some new information about the future value of the stock.)
If there’s a distribution over who has the market-beating analysis on any given day, then with some probability, it’s you. So the question is, what are the odds?
It’s not going to be a uniform distribution over people or predictions. There’s some prior based on what information you have access to, and how good at certain kinds of reasoning you are compared to others, and then you update based on how strong your inside view is. (And over time, you can update on track record.)
Most of the time, it will probably be professionals who do this for a living. But it’s not crazy to me that as an intelligent amateur who consumes lots of information and thinks carefully, it will sometimes be your turn, so to speak.
Btw, it seems like there’s a little bit of a paradox here. Suppose the participants in a market are all well-calibrated about whether they have market-beating information. And in expectation everyone will stumble upon some market beating information at some time or another.
On the one hand, everyone should be able to get market returns, by just investing in the market. But since everyone has market-beating information some of the time (and are well-calibrated), they should actually be able to beat the market, by diverging from the market portfolio only when they have market-beating information.
But then you’d have everyone beating the market, which can’t be possible.
I think the resolution is the same as Zvi’s point in the Sources of Disagreement and Interest (aka Suckers at the Table) section of Prediction Markets: When Do They Work? In the scenario I describe, no one would be able to profit off of their private information. As soon as they go to buy on positive information (or sell on negative information), they’ll find that everyone else (who’s well-calibrated) raises (or lowers) their prices to exactly match. (Or that no one was showing any bids or offers in the first place, since they knew any fills would face adversarial selection.)
In reality, I think the way this works is that you have both suckers (people who are miscalibrated), and also people who are willing to go forego market returns—that is, people who have some natural reason to buy (e.g. new investors) or sell (e.g. in order to spend).
Suppose the participants in a market are all well-calibrated about whether they have market-beating information.
That part seems particularly unrealistic. If that were true, we’d be living in a very different world.
Many large market participants have perverse incentives when trading other people’s money. Their customers would prefer low volatility over optimal Kelley bets, and many would have to panic sell in a drawdown if volatility wasn’t kept under control. You don’t have to be smarter than them to exploit them, since they’re optimizing a different goal: keep their customers happy, instead of making maximum money for them.
And then, most people are irrational. Going by base rates, you should expect other market participants, even big ones, to trade emotionally as well.
You don’t have to be smarter than them to exploit them, since they’re optimizing a different goal: keep their customers happy, instead of making maximum money for them.
Sell puts when implied volatility (IV) is higher than usual, on stocks where the IV tends to exceed the historical volatility (this is most of them, actually).
Whales have to buy puts for more than they’re really worth to protect their customers’ portfolios from scary market volatility.
Buy them back for less than you were paid for them when IV reverts to the mean. It’s like selling insurance. You have to control your bet size and hedge (maybe with a cheaper put, like reinsurance) so you don’t get wiped out when the disaster actually happens, but you’ll get more than enough premium to make up for your losses.
I think it is very relevant to note that you were up 50x at one point and then down to 10x on net after further decisions went sour, because
those further decisions going sour show your decision-making was not that consistently good
such high variance looks much more obviously like ‘gambling’ or ‘taking on an enormous amount of risk’ than ‘it’s fun and easy to seek out alpha and beat the market’
such high variance looks much more obviously like ‘gambling’ or ‘taking on an enormous amount of risk’ than ‘it’s fun and easy to seek out alpha and beat the market’
I know someone else who made the opposite mistake as me and sold their coronavirus puts too early. If you only saw their record, there would be no “high variance”. They just made less money than they could have. It seems to me that the correct lesson from both outcomes is that it’s possible to beat the market (without putting in so much effort as to make it not worthwhile to try), but we haven’t figured out how to time the exits at exactly or very close to the best times.
When Eliezer posted about my bet, I had only gained 7x my initial bet (and that’s what he posted about), and although I ended up losing 80% of my paper profits (which were 50x at one point) I still gained 10x my initial bet. So him not posting further followup seems fine? And at least from my personal perspective (i.e., where selection bias isn’t an issue) the final outcome still seems to be strong evidence against EMH.
It seems to me that the EMH is a spherical cow. In reality, there has to be some process by which market prices comes to reflect new information. And whoever figures it out first, either because they had private information, or because they did better analysis, should make some profit.
And the question might be—is it always the same people who figure it out first, or is there some randomness involved? (Consider a toy model where N people are trading one stock and every day a fairy randomly selects 1 of the N people and tells them some new information about the future value of the stock.)
If there’s a distribution over who has the market-beating analysis on any given day, then with some probability, it’s you. So the question is, what are the odds?
It’s not going to be a uniform distribution over people or predictions. There’s some prior based on what information you have access to, and how good at certain kinds of reasoning you are compared to others, and then you update based on how strong your inside view is. (And over time, you can update on track record.)
Most of the time, it will probably be professionals who do this for a living. But it’s not crazy to me that as an intelligent amateur who consumes lots of information and thinks carefully, it will sometimes be your turn, so to speak.
Btw, it seems like there’s a little bit of a paradox here. Suppose the participants in a market are all well-calibrated about whether they have market-beating information. And in expectation everyone will stumble upon some market beating information at some time or another.
On the one hand, everyone should be able to get market returns, by just investing in the market. But since everyone has market-beating information some of the time (and are well-calibrated), they should actually be able to beat the market, by diverging from the market portfolio only when they have market-beating information.
But then you’d have everyone beating the market, which can’t be possible.
I think the resolution is the same as Zvi’s point in the Sources of Disagreement and Interest (aka Suckers at the Table) section of Prediction Markets: When Do They Work? In the scenario I describe, no one would be able to profit off of their private information. As soon as they go to buy on positive information (or sell on negative information), they’ll find that everyone else (who’s well-calibrated) raises (or lowers) their prices to exactly match. (Or that no one was showing any bids or offers in the first place, since they knew any fills would face adversarial selection.)
In reality, I think the way this works is that you have both suckers (people who are miscalibrated), and also people who are willing to go forego market returns—that is, people who have some natural reason to buy (e.g. new investors) or sell (e.g. in order to spend).
That part seems particularly unrealistic. If that were true, we’d be living in a very different world.
Many large market participants have perverse incentives when trading other people’s money. Their customers would prefer low volatility over optimal Kelley bets, and many would have to panic sell in a drawdown if volatility wasn’t kept under control. You don’t have to be smarter than them to exploit them, since they’re optimizing a different goal: keep their customers happy, instead of making maximum money for them.
And then, most people are irrational. Going by base rates, you should expect other market participants, even big ones, to trade emotionally as well.
What trades does this suggest?
Sell puts when implied volatility (IV) is higher than usual, on stocks where the IV tends to exceed the historical volatility (this is most of them, actually).
Whales have to buy puts for more than they’re really worth to protect their customers’ portfolios from scary market volatility.
Buy them back for less than you were paid for them when IV reverts to the mean. It’s like selling insurance. You have to control your bet size and hedge (maybe with a cheaper put, like reinsurance) so you don’t get wiped out when the disaster actually happens, but you’ll get more than enough premium to make up for your losses.
I think it is very relevant to note that you were up 50x at one point and then down to 10x on net after further decisions went sour, because
those further decisions going sour show your decision-making was not that consistently good
such high variance looks much more obviously like ‘gambling’ or ‘taking on an enormous amount of risk’ than ‘it’s fun and easy to seek out alpha and beat the market’
I know someone else who made the opposite mistake as me and sold their coronavirus puts too early. If you only saw their record, there would be no “high variance”. They just made less money than they could have. It seems to me that the correct lesson from both outcomes is that it’s possible to beat the market (without putting in so much effort as to make it not worthwhile to try), but we haven’t figured out how to time the exits at exactly or very close to the best times.