Poker-AI.org

Hi,

There is lots of way to make a poker bot. You can :
- Use some advanced algo,
- Use your own poker knowledge,
- Use a GTO approach .... .

With time, lots of GTO solver are created and I wonder if someone already try to use it inside his bot.
Only thing you need to do it's get all information for the calcul but the major problem is the calculation time I guess.

I'm curious to know if someone already think about this solution.
Thanks

Yes

People do this, I don't think it's the best idea though.

The problem is, that you can use a good abstraction, so it will all be much bigger than it has to or you will have to allow only a very small number of actions.

I am starting to use solver data in certain spots that I have difficulty or uncertainty with. Right now just SB UOPFR preflop. I would use Solver data for any play on the blinds. I was way off on correct SB strategy so Solvers are really useful and a necessary tool.

Update: I started using solver data for my entire preflop strategy. These tools are a must have and will change the game. In the next few weeks I plan to implement a post flop strategy and will see how it goes.

Anybody knows, which exactly algorithm used in solvers (PIO, CREv, etc) ?
Modifications of CFR?
Are there other algorithms open source for building equilibrium?
If you have finalized the CFR, do the test results coincide with the solvers?

I'm ready to buy such a modified algorithm, which can repeat CREv tests

CFR is the unique solver suitable to a poker game. There're bots GTO based whose win in differents poker-room, i know it because i was a PPP (professional poker player) and looking at the hands of my friends I can recognize them. Anyway i think that most of these bot are based on reinforced learning like snowie becouse using cfr in games to more than two players is not straightforward and is not said to be a good choice.

CFR is a reinforcement learning algorithm