1

I'm a programmer with a background in mathematics, but I have no experience whatsoever with artificial intelligence/neural networks. I'd like to study it as a hobby, and my goal for now is to solve the following simple poker game, by letting the program play against itself:

We have two players, each with a certain number of chips. At the start of the game, they are obligated to put a certain amount of chips in the pot. Then they each get a random real number between 0 and 10. They know their own number, but not the one of their opponent. Then we have one round of betting. The first player puts additional chips in the pot (some number between 0 and their stack size). The second player can either fold (put no additonal chips in the pot, 1st player gets the entire pot), call (put the same number of chips in the pot, player with highest number gets the pot) or raise (put even more chips in the pot, action back on 1st player). There is no limit to the amount of times a player can raise, as long as he still has chips behind to raise.

I have several questions: - Is this indeed a problem that can be solved with neural networks? - What do you recommend me to study in order to solve this problem? - Is it feasible to solve this game when allowing for continuous bet/raise sizes? Or should I limit it to a few options as a percentage of the pot? - Do you expect it to be possible to get close to an equilibrium with one nightly run on an 'average' laptop?

Although not a direct answer, the problem you have is a kind of minimax search. There has been a lot of research on these types of games which means its a great place to start! These kinds of problems have been approached using monte carlo tree search which I highly recommend reading about as it should help guide your questions: https://int8.io/monte-carlo-tree-search-beginners-guide/

– Jaden Travnik – 2019-01-16T22:11:37.990