ReWeighting(Action a, WeightTable wt, GameContext gc, OppModel om){
for each (hand in wt){
ProbabilityDistribution pt[FOLD,CALL,RAISE];
PT = predictOppAction(om, hand, gc);
wt[hand] = wt[hand] * pt[a];
}
}
but I found a problem, in predictOppAction i need calculate EHS with hand+gameContext.
My source and my processor let me get the flop EHS in 3 seconds, and 3 seconds for each entry in WeightTable is so much... I thought save in a data file, all EHS of all possible context, but this would, 6 billions of seconds, and indexing in 2 billions data, for flop only, is not nice..
I read about LUTs but i don't understand, I read about reduce the possibilities, by similarity patterns, but calculate EHS of 1 case, which represents 100 cases,is not very accurate...
Did you use this pseudocode?
there any other way to do it??
Thanks...Statistics: Posted by karma — Sun Mar 01, 2015 6:29 pm
]]>