Poker-AI.org

That highly depends on the knowledge of the poker guy and his ability to generalize his thoughts. Too much knowledge might actually be hindering - even if the edge was greater

this was enlightening for me! ty, regular member!

Yeah the problem is that the fine tuning takes the longest to do as the little things make the biggest difference.




As a whole its not really complex (not like MCTS). Getting the range correct pre for me is the biggest hurdle as it obviously affects the rest of the hand and if your wrong it could really cost some big EV. As I mentioned previously the biggest problem is the bluffing ranges of the villain. Villains can have identical stats but completely different bluff ranges which has caused me some big headaches. Generally it seems most decent players will 3B bluff hands that can take a call and not just total trash. So SC's, blockers, etc.

Post flop is fairly straightforward and it comes down to basic hand combinatorics and range reduction. So of his range what % is value, air, draws. And if he is not leading the betting then what part of his range can be called etc. On each street its going to get tighter and tighter as the range he can bet for value gets smaller and smaller until your left with a very small number of hands that he could possibly reach the river with. So on the river your left with value, bluff catchers and missed draws. So now its just a matter of running his range vs hand and its gets easier to figure out what to do..either to bluff catch, bluff or value bet. But..its not just a matter of equity but also a matter of hand combinatorics to figure out how to extract max value. Like if his range is mostly missed draws you would be way better off letting him try to bluff at you vs leading out. You get the idea..

Hey guys, we are still working on the thing and now I am getting a programmer to work on the "stealth" part of the project.

He is suggesting that for our setup we should use two different physical machines, with one of the two grabbing poker data from the client via OCR, but he says that the OCR inaccuracy in number recognition can be a big problem (words, not much so as we have a very small vocabulary in poker).

I have noticed thou that this isn't accounted as a big problem in general by the botting community, how is that so? Am I getting something wrong here? Maybe that is because most people do the DLL injection thing and or similar and not the OCR trick?
Thank you!