Different stack and blind sizes about deepholdem HOT 4 CLOSED

I think the bad solution you mentioned is actually a good solution.
Because the models are trained with lots of varying pot sizes, it can handle any stack size between 1bb and 200bb. If you scale the pot and stack so that stack = 20000 chips, then you can input the situation into the model, as you said.

I don't know how deepstack did it, but I think they will use this method too.

from deepholdem.

Ah that's great. Thank you, that makes a lot of sense. I agree with you, deepstack said they did not retrain their models. So there really is not another way of doing it other than the way you described.

@happypepper This is a bit offtopic but I am really interested what projects you are currently working on. Obviously you dont have to say if you dont want to. Id also be interested in your plans for future poker projects. I personally think 6 man / 9 man poker would be really challenging and interesting.

from deepholdem.

@PhDtimothy and @happypepper I was wondering how exactly one would implement this in code. I have changed the code so I can simply input the current matchstate as a string into deepstack.lua and it outputs its action. But I am not exactly sure how I can change the blind size properly after a hand. Would it be enough to change the arguments.bb and arguments.sb size after a hand is played, so that deepstack.lua solves for the new blind sizes? Thank you many times for your help.

from deepholdem.

Hey @Jedrickbender , did you be able to resolve it? I want to test dynamically different stacks and I'm having issues

from deepholdem.