This project is aimed at studying the use of Machine Learning in Strategy games like chess
We have developed 2 chess-like games with new mechanics and pieces that have never been statistically analysed in the past. These games are going to be the target of the computer-player as it will play itself and analyse the game for move patterns, etc. This will first be tested on a chess interface before being exported to the other 2 custom games nicknamed - Magni and Modi.
The ultimate aim of the bot is to try to solve the new games we have designed. A game is considered solved if its outcome can be determined correctly given any game position (assuming all players play their best moves, i.e. they play perfectly). In this respect, Tic-Tac-Toe, checkers and Connect-4 are considered solved games. However, games like Chess and Go are considered extremely complex, and have so many possible states and moves till date, these games remain unsolved. The best we do with them is to calculate the win probability of each player at any point in the game. This is why computer chess and go players cannot definitively play better against a human player. To solve a game answers many questions about it, including whether or not it is balanced (whether the starting position gives any player an advantage from the get-go)
These statistics are ultimately decided by games played and their outcomes. Our aim with this project is to use Machine Learning to recreate this data and recognize move-patterns to be favoured in various situations.
- Typescript, JS
- CSS
- HTML