simple kotlin chess engine

Simple Kotlin Chess Engine, or just kengine, is a minimalistic chess engine written in kotlin. This project aims at demonstrating how to build a chess engine using a chess library and also at providing a basic project for people interested in quickly bootstrapping a chess playing agent for easily adding and test his/her own ideas.

Building

From source

$ git clone git@github.com:bhlangonijr/kengine.git
$ cd kengine/
$ ./gradlew clean build

Usage

kengine can be used in a chess GUI that supports the UCI protocol by implementing the basic commands that communicates with it. Refer to UCI documentation to have a better understanding on how it works.

Using the command-line interface

# Look for a kengine.sh or kengine.bat in the root folder and execute it
$ ./kengine.sh

Type uci and hit enter to display engine info.

$ uci 
id name kengine 1.0.2
id author bhlangonijr
option name Hash type spin default 128 min 1 max 16384
option name Threads type spin default 1 min 1 max 128
option name SearchAlgorithm type combo default MonteCarlo var AlphaBeta var MonteCarlo
uciok

uciok means the program is ready to accept commands.

Example 1: Increasing the hash table size to 256mb

$ setoption name Hash value 256

Example 2: Calculating the best move for the initial standard position (limiting search to depth 5)

Type go depth 5 and hit enter:

$ go depth 5
info depth 1 score cp 50 time 19 nodes 24 nps 24 pv b1c3
info depth 2 score cp 0 time 48 nodes 131 nps 131 pv g1f3 g8f6
info depth 3 score cp 50 time 69 nodes 787 nps 787 pv g1f3 g8f6 b1c3
info depth 4 score cp 0 time 128 nodes 4437 nps 4437 pv g1f3 g8f6 b1c3 b8c6 g5h7
info depth 5 score cp 40 time 195 nodes 22791 nps 22791 pv g1f3 g8f6 b1c3 b8c6 e2e4
bestmove g1f3
info string total time 195

Installing the engine in a chess GUI

For playing with kengine using your preferred chess GUI it only takes providing the startup script kengine.sh or kengine.bat in your chess GUI UCI engine setup options.

Self-play using integration tests

kengine comes with two different bare bones Search algorithms which can be set via options: Alphabeta and Monte Carlo. Integration tests in the project test folder can match one search against the other:

    @Test
    fun `match Abts and Mcts engines`() {

        val board = Board()
        board.loadFromFen(startStandardFENPosition)

        val abts = AlphaBetaSearch()
        val mcts = MonteCarloSearch()

        val moves = MoveList(board.fen)
        while (!board.isDraw && !board.isMated) {
            val move = play(board, abts, mcts)
            if (move != Move(Square.NONE, Square.NONE) && board.doMove(move)) {
                moves += move
                println("Played: $move = ${board.fen}")
            }
        }
        printResult(moves, board)
    }

Features

UCI protocol support;
Basic time control;
Alphabeta search algorithm with Null move pruning, PVS (principal variation search), transposition table and quiescence search;
Basic Monte Carlo search with UCT;

bhlangonijr / kengine