Loading…
Glossary term
A search method that evaluates possible moves by running many random simulations, used in game-playing AI.
MCTS decides on a move by building a search tree guided by many random simulations. It repeatedly selects a promising branch, expands it, plays out a simulated game to the end and updates the statistics back up the tree. Combined with neural networks, it powered AlphaGo.
It lets programs plan in huge search spaces like Go without exploring every possibility.