Skip to main content
edited tags
Link
manlio
  • 4.3k
  • 3
  • 28
  • 38
Source Link
Amja
  • 113
  • 4

When implementing Monte Carlo Tree Search for TicTacToe, do I simulate won states?

I'm trying to debug my MCTS implementation for TicTacToe (it doesn't block obvious wins for the opponent). I was wondering what the algorithm should do if it expands to a node which is a game over state. Should it continue to "simulate" that node and back-propogate the results up the tree or just ignore it if the node is chosen.