Backend side of the learning chess application
- Java 8
- Maven
- PostgreSQL as database
You need lchessrulz installed on your local maven repository (it is not part of maven central repository)
Checkout this repository and then run mvn clean install to install it.
Create a learning-chess database on your PostgreSQL server.
To run the project from Eclipse:
- Import the project into eclipse as maven project
- Locate ChessApplication and run it as a Java Application.
From the command line use:
java -jar target/learning-chess-back-0.0.1-SNAPSHOT.jar
This is a backend side of the learning chess application. Communication from the frontend happens through REST. This is programmed with Java 8 and leans heavily on Spring Boot.
Learning chess utilizes past games when deciding what move to make. In theory, it should be able to defeat or get a tie against any opponent given enough games and times. To test this, I created a relatively simple sparring AI partner for this and made them play games against each other in loop. It took approximately 3000 games before learning chess either defeated or got a draw against it with regular basis.
Basically, when the game is completed and it was not a draw, every board state of every move will be stored into the database. If there is no previous entry for the board state, then a new one is created with score of 1 or -1 depending whether the winner was white or black player. If board entry exists already, then 1 or -1 is added to the existing score.
First learning-chess will create moves for the given board state and then fetch scores for those moves from the database defaulting to 0 score for the move if no entry is stored. Learning chess will then pick a move with highest value for white player or lowest for black player. If there are multiple moves with the same score a simple evaluation method is applied for board state to pick a move from those.