Poker-AI.org

Has anybody done any work on equity realization? I am in the midst of doing some basic estimation of it but its far from accurate. There are so many factors that come into play that I am not sure if it can really ever be figured out but I had to come up with something. The EV calcs can be way off because of this factor.

My first idea was to take the frequencies of facing a bet on the turn and river and then figure out how often I would be calling on average. Lets say its 43% on the turn and 15% on the river. In general I would be calling around 66% of time if I faced a half pot bet (I was thinking of running some queries on the limits the bot is playing to find the average bet size it faces on each street). This is because I would never be folding more then 33% of the time as the villain would be getting value on all of his bluffs. So then you can multiply those percents together to get an average folding percentage: 9.9*28 = 27%. So on average I would be realizing 73% of my equity. Mulitway this is probably ok but for HU its probably much higher as I would not be facing a triple barrel that often.

One problem is that different equities have different realization rates. Obviously we are going to realize more of our equity if we flop the nuts and less if he have two pair. The nut type of hands usually play themselves so they are not too concerning. The ones that concern me are the equities between 33-50% which represent a huge chunk of range.

The only idea I can come up with is running sims on HH data. An example is taking a 39% equity hand find out how often it reaches showdown for HU, 3 way, 4 way etc. Not so trivial to do but all I really want is something in the ballpark of correct.

Any insights on this problem would be greatly appreciated.