Poker-AI.org

Hi!

My team has fully undetectable by PokerStars bot system (software+hardware) and rule-based winning algo for micro limits (no limit cash). So we have all needed experience in botting plus all needed environment.
We are looking for help with AI-based algo (cash or sng). If you have any thoughts (or ready to run algo) and ready for cooperate - welcome to my skype: babushkamoroz

up!
Don't be shy - pm me anywhere!

Finally we have found it, only 1 year later! https://www.mypokersolver.com/

Your wasting your time on 6max solver simulations. They are break even at best and -EV without rakeback. I know this from experience. They are good for basic strategy and training but that is about it. We are still a few years away from solving post flop 6 max with NL holdem being worse then PL Omaha. For preflop solvers are great but for post flop everything has to be run independent as the frequencies vary WIDELY.

Yes, preflop is very good. Already implemented. About postflop for 6 handed not perfect with deep stack your right, but short stack strategy is very helpful.

The issue with gto strategy is that it is just good for practicing how should solid unexploitable ranges look like. In online poker, all poker sites are full of collusion, team play, so bots that use gto strategy are just donating money. Even if the game would be fair, rake is too high, gto can't beat the rake you need to exploit opponents.

I am not even sure if 50bb post flop is that helpful tbh. In general the preflop simulations make post flop severely passive. The number of problems are endless..one is post flop bet sizing. Your going to have to make rules to handle all the different possible sizings. Most 6max solver simulations only have one or two post flop sizings which is not even close to enough. You need at least 4 with another 3 raise sizings (and this is just for PLO). All of these sizings are board/range dependent so its not so simple.

OOP and 3B pots is where your going to see the biggest problems. Your going to notice heavy losses on the blinds in SRP and you will struggle to win back half the blinds with an all solver strategy. The other problem is that solvers are terrible vs weak players on later streets. Your going to call that river bet at GTO and your RCE is going to plummet because weak players do not bluff the river.

I have been running solver bots since early this year and it was a waste of time but very enlightening. It shows signs of brilliance but in general it was ZERO EV. I think it could do ok with decent rakeback. Rakeback solves issues with a weaker AI and in lower limits the all solver strategy IS A WEAK AI in its current form for all 6max games.

Depending on the limits your playing the rake is also a huge factor on the preflop strategy. I was really shocked to see the preflop frequency differences especially on the blinds. And..lets not forget you have to ditch the solver entirely in limped pots and your going to have to think on your own. If you want a puzzle to figure out try finding the correct over limp strategy for the SB. I worked on that for over a year until I understood the problem correctly. It looks simple but its not.

I could go on and on.

The Libratus match demonstrates that gto is more than adequate for HU.

I have been doing mostly HU stuff in the last few months and I can attest that GTO is good for HU without a doubt. Adding in exploitative plays is of course very beneficial in certain spots. As with 6max calling on the river vs a weak player with a marginal hand will -EV long term.