macrocephalopod

A nice mental model in poker is to measure your profits in “Sklansky dollars” instead of real dollars. A Sklansky dollar is the expected value from a play if you knew what every other player held.

For example, you hold AA vs KK pre-flop and your opponent goes all in for $100. If you call you have about an 83% chance of winning a $200 pot, so your Sklansky dollar profit from calling is 83% x $200 - $100 = $66

Clearly you will either lose or gain $100 depending on how the cards fall, but that’s just luck. Measuring success in Sklansky dollars instead of real dollars removes an element of chance from the game, making performance measurement more reliable.

If your per-hand profit in real dollars is going up, that could be because you’re getting better but it could just be luck. If your Sklansky dollar profit is going up, that’s likely to be because you’re getting better.

Sports betters can use a similar heuristic. Instead of measuring profit in real dollars, assume the odds immediately before the start of the game are accurate, and measure your profit in terms of how much the odds moved in your favour from when you bet to when the game started.

By removing the randomness from the game from your measurement, you can zero in much faster on your true edge and calibrate your bet size and frequency more appropriately.

An application in finance is transaction cost analysis. If you measure execution slippage vs arrival price, it’s very noisy. Instead measure it vs a VWAP over your execution window, or if measuring slippage in single names, adjust the benchmark to take account of how much the …

…index or sector moved over the execution window. There are similar techniques for measuring the quality of your alphas.