I should have known the Quant Trading X Poker intersection is large. Since there is some engagement here, let me share something interesting that is relevant to both sides of the house.
Non-ergodicity. If you google this word it will tell you something about something not visiting all states of the system. This is cool and all, but what it's really about, is that your EV calculations are flawed because they are predicated on a large (infinite) sample size, and in a non-ergodic system, there are various impediments to reaching that large (infinite) sample size.
Poker as a game is non-ergodic in the way that markets are - your wealth is not infinite, and you can get stopped out before you reach your expected value. Good risk-takers intuitively understand this, and this is why newer risk-takers will receive "wisdom" that take the following form:
1. Start small and start with small vol.
2. Get some PnL before you upsize / take bigger risks
Conventional wisdom is often a heuristic for something complex because memes travel and terse writings do not. The heuristic here is simple - survive long enough so that the expected value of your strategy can be realized.
This is done by not getting fired (stopped out) and not putting yourself down on a path where you cannot recover. Realize that if you lose 50% of your wealth, you need 100% returns on the remaining wealth to get up to the starting line. If you lose yet another 50%, you will need 400% returns on the remaining wealth to get up to the starting line. I think this is known either as the logarithm of death or logarithmic hell in Mark Spitznagel's book. It is a kind of path dependency that all but ensures you get stopped out.
Some situations are "more non-ergodic" than others. You could roughly argue that cash games in Poker are ergodic because you can always borrow more to buy-in (up to a limit of course), but put yourself in a poker tournament and it is an EXTREMELY non-ergodic system. Likewise, in markets, if you are an independent trader, you could argue that you are operating within an ergodic system, but being a portfolio manager in Millenium would be a highly non-ergodic system since a feather would cause you to stop out and get fired.
So... all is this to say... Expected value calculations are interesting and all, but if you are not able to play a large enough sample size to realize those expected values, they become inherently meaningless and flawed.
Non-ergodicity. If you google this word it will tell you something about something not visiting all states of the system. This is cool and all, but what it's really about, is that your EV calculations are flawed because they are predicated on a large (infinite) sample size, and in a non-ergodic system, there are various impediments to reaching that large (infinite) sample size.
Poker as a game is non-ergodic in the way that markets are - your wealth is not infinite, and you can get stopped out before you reach your expected value. Good risk-takers intuitively understand this, and this is why newer risk-takers will receive "wisdom" that take the following form:
1. Start small and start with small vol.
2. Get some PnL before you upsize / take bigger risks
Conventional wisdom is often a heuristic for something complex because memes travel and terse writings do not. The heuristic here is simple - survive long enough so that the expected value of your strategy can be realized.
This is done by not getting fired (stopped out) and not putting yourself down on a path where you cannot recover. Realize that if you lose 50% of your wealth, you need 100% returns on the remaining wealth to get up to the starting line. If you lose yet another 50%, you will need 400% returns on the remaining wealth to get up to the starting line. I think this is known either as the logarithm of death or logarithmic hell in Mark Spitznagel's book. It is a kind of path dependency that all but ensures you get stopped out.
Some situations are "more non-ergodic" than others. You could roughly argue that cash games in Poker are ergodic because you can always borrow more to buy-in (up to a limit of course), but put yourself in a poker tournament and it is an EXTREMELY non-ergodic system. Likewise, in markets, if you are an independent trader, you could argue that you are operating within an ergodic system, but being a portfolio manager in Millenium would be a highly non-ergodic system since a feather would cause you to stop out and get fired.
So... all is this to say... Expected value calculations are interesting and all, but if you are not able to play a large enough sample size to realize those expected values, they become inherently meaningless and flawed.
I believe this 50-year-old system is the prevailing model for dealing with non-ergodicity in poker tournaments: en.m.wikipedia.org
Loading suggestions...