Quote:
PM me your algo for computing best response/exploitabillity in imperfect-recall games/abstractions and maybe we can work something out
Poker is a perfect recall game and that you might use imperfect recall abstraction to approximation NE for that doesn't matter. As long as you can play the game you can calculate max exploit. Just travel the tree in depth first fashion, go to every leaf (you can do that if you can play the game), calculate stuff there (pot equity vs whole range), go back to root (while making calcs for both max exploit and ev), profit.
It may take a while for a whole game but it will be relatively fast for a flop. If the abstraction is so huge it won't be fast on the flop (like you know, you assume every size is possible) then cut the betting abstraction to like 3 sizes (for max exploit not for the bot) and calculate this way (so max exploit assuming that max exploit can use only 3 sizes).
The fact that you play flops/turns instantly suggest you have that stored. If that's the case then it really isn't a big computational challenge to do so. You can even easily split it to as many cores as you want.
Quote:
I object to this post. You are implying that I stated all the things you are saying. E.g., "you claim stuff that you have this amazing thing and people can learn a lot of from it then provide the number quantifying how good it is." But in post http://forumserver.twoplustwo.com/29...postcount=1070 you only give links to quotes by Professor Sandholm, and you never clarified whom your different "the author"'s refer to. Please stop attributing quotes to me that I did not say.
By "you" I mean the team and by the "author" I referred to him (as he appears to be the main man behind the project) because he is the one which appears to be someone without much knowledge specifically about poker and making big claims about the bot.