b-inary/postflop-solver

node lock

tankche1 opened this issue · 2 comments

tankche1 commented

Any plan to support node lock? It looks straightforward from the current code. Specifically, can we call the solve function after we run some game.play(), and then it will start searching from the current state? This may need some modification in solve_step to determine the current player and starting node.

mhluska commented

If it's straightforward, why don't you submit a PR?

  • 😄1
b-inary commented

I wouldn't say the implementation is complete, but this engine repository actually already supports node locking. It supports normal node locking (test code), as well as partial range node locking (test code).

I also note that you seem to misunderstand the principle of node locking. Node locking can affect nodes before the locked node. For example, if we know that the opponent will always fold after our bet (i.e., the opponent's future node is locked), then we would increase our bluff bets. This is completely different from just solving with different ranges.