chaitanya100100/AI-Agent-for-Ultimate-Tic-Tac-Toe

An AI agent implemented using Monte Carlo Tree Search (MCTS) using Upper Confidence Bounds (UCT).

AI-Agent-for-Ultimate-Tic-Tac-Toe

An AI agent implemented using Monte Carlo Tree Search (MCTS) using Upper Confidence Bounds (UCT).

Game Description

• Basic game is - https://mathwithbaddrawings.com/ultimate-tic-tac-toe-original-post/
• But instead of 3*3, it is 4*4

Resources for MCTS and UCT

• https://en.wikipedia.org/wiki/Monte_Carlo_tree_search
• https://jeffbradberry.com/posts/2015/09/intro-to-monte-carlo-tree-search/
• http://www.cameronius.com/research/mcts/about/index.html

Test

• Participated in AI tournament() in which total 90 teams participated. Qualified for semifinals along with 30 teams.
• Semifinals had 3 pools each having 10 teams. Secured 5th rank in semifinals.

Scope of improvement

• Backpropogation step of MCTS properly to store heuristic
• Implementation in C++ instead of Python for more number of simulations
• Evaluation function