Gambler's Reward: Optimal Betting Size

post by b1shop · 2012-01-17T20:32:31.003Z · LW · GW · Legacy · 10 comments

Contents

10 comments

I've been trying my hand at card counting lately, and I've been doing some thinking about how a perfect gambler would act at the table. I'm not sure how to derive the optimal bet size.

Overall, the expected value of blackjack is small and negative. However, there is high variance in the expected value. By varying his bet size and sitting out rounds, the player can wager more money when expected value is higher and less money when expected value is lower. Overall, this can result in an edge.

However, I'm not sure what the optimal bet size is. Going all-in with a 60 percent chance of winning is EV+, but the 40 percent chance of loss would not only destroy your bankroll, it would also prevent you from participating in future EV+ situations. Ideally, one would want to not only increase EV, but also decrease variance.

Objective: Given a distribution of expected values, develop a function that transforms the current expected value into the percentage of the bankroll that should be placed at risk.

I'm not sure how to begin. Even if I had worked out the distribution of expected values. Are other inputs required (i.e. utility of marginal dollar won, desired risk of ruin)? Should the approach perhaps be to maximize expected value after one playing session? Why not a month of playing sessions, or a billion? Is there any chance the optimal betting size would produce behavior similar to the behavior predicted by prospect theory?

I eagerly await an informative discussion. If you have something against gambling, just pretend we're talking about how much of your wealth you plan on investing in an oil well with positive expected value.

10 comments

Comments sorted by top scores.