Poker-AI.org

Hi,

I created a rule based 6-max bot where the rule parameters are optimized by evolution with playing against itself.

The funny thing is that I cannot beat my very first completely stupid approach1 (where the bot had no limits what it does preflop).

The problem is that I use the same very simple postflop play in my newer bots too because I wanted to know what is the best preflop approach. However, preflop and postflop play are connected (who would have thought). My postflop play only uses hand strength, PPot, position, numraisers, potSize, toCall and a few optimizable parameters to decide what to do.

The strategy that hurts watching but I cannot beat is that my stupid approach evolved to only bluff with the very worst preflop hands. That makes postflop play much easier because if called it knows it lost.

My current preflop play (approach4) is very simple but I thought smarter, but it has no chance. For certain x% top cards it raises, y% top cards it calls, modulated by position, number of raisers and pot size. I assumed this would lead to bluffing with "good cards" (so x large for 0 raisers late position cases).

So I think I have to create a more complex postflop play to be able to beat that. Or does anybody have a simple idea what to add to my current approach to beat the stupid one? I already tried different distributions for what x% top cards means, but I cannot beat the stupid approach.

Thanks and regards

Hello,

I prefer Selby's preflop Game in HU.

More Infos here

Raising pre-flop

http://poker-ai.org/phpbb/viewtopic.php?f=26&t=2666


or e.g 'hutchinson poker table' for 6 Player or more.

http://www.online-gambling.co.uk/articles/Hutchinson-strategy-system.asp


Tom

Thanks, but I think my approach is more flexible than hardcoding stuff.

I found the problem and now my new bots have no problems destroying the old ones.
Had two ugly programming errors in the current approach (I apply curves to some parameters, and they switched signs in the middle because of a sign error in the formula. And my bots thought they are committed to the hand much too often because of my stupidity multiplying a normal distribution by 5 instead of 0.05. A wonder that those bots played anything at even looked like poker).