This repository is a collection of chess strategies, as well as a small framework for playing against them in the python interactive mode.
These strategies are largely inspired by Tom 7's various chess shenanigans. After watching his Elo World video, I wanted to play against some of the chess players myself, but didn't feel like figuring out how to compile his C++ code and get all that working. So rather, I decided to reimplement parts of it in Python and use nice pre-existing libraries to do the heavy lifting.
| Name | Description | Properties |
|---|---|---|
| RandomMove | Just picks a move at random | |
| MinResponses | Picks a move that minimizes the amount of responses that the opponent will have | |
| SuicideKing | Tries to move its king as close as possible to the enemy king | |
| Stockfish | Plays moves from the Stockfish engine, by default at depth 18 | Engine |
| GnuChess | Plays moves from the GNU Chess engine in UCI mode, by default at depth 10 | Engine |
| Worstfish | Uses Stockfish at depth 10 to play the worst possible move | Engine |
| LightSquares | Prefers moving its pieces onto light squares | |
| DarkSquares | Same as above but for dark squares | |
| Equalizer | Will move the piece that has moved the least number of times onto a square that has been visited the least number of times. Fewer moves takes priority over less visited squares | Stateful |
| Swarm | Moves pieces as close as possible to its enemy's king | |
| Huddle | Moves pieces as close as possible to its own king | |
| LightSquaresHardMode | Prefers moving its pieces onto light squares, but will let Stockfish choose which move to play out of the moves that maximize light squares | Engine |
| DarkSquaresHardMode | Same as above but for dark squares | Engine |
| Possessed | Uses Stockfish, but once a piece has moved, that piece is "possessed" and will continue to be moved until it is captured or has no legal moves | Engine, Stateful |
I've tested this with python3. It might work with python2, idk ¯\_(ツ)_/¯
The only library needed is python-chess:
pip install chess
To use engines: Make sure that stockfish and gnuchessu are installed and on your PATH. Stockfish can be downloaded on Windows, Linux, and macOS from their official website. GNU Chess can be installed on Linux via your favorite package manager, or built from source to get the latest version. There might be binaries for other platforms available for download ¯\_(ツ)_/¯
Launch python in interactive mode:
python3 -i main.py
Now, you can set whatever strategies you want by setting the strat1 and strat2 variables:
>>> strat1 = MinResponses()
>>> strat2 = Worstfish()
Finally, run a game:
>>> run_game()
The game will be printed in PGN format as it unfolds.
To play against the strategies: Make one (or both) of the strategies an instance of Human. For example:
>>> strat1 = Human()
>>> strat2 = LightSquares()
>>> run_game()
When it is your turn, it will ask for your moves which you can enter in SAN or UCI format.
Using uci-ify.py, you can run a strat as a simple UCI engine! To use a specific strat, either run with a strat name as a parameter:
./uci-ify.py Swarm
Or, set the option Strategy to the strat you want to use in your UCI-compatible GUI of choice.
Note on stateful strategies: Stateful strats are interesting because the UCI protocol is inheriently stateless. Each move is sent via a new position command and that sets the internal state of the board - moves are not built on top of previous moves. For that reason, stateless strats are basically reset on each position command, and then the entire game is replayed to rebuild that strat's state.
As I stated earlier, these strategies are inspired by Tom 7, and some of those strategies were very interesting (such as the blindfolded strats), and I want to get as many of those implemented as possible.
Here are some things I'd like to get done at some point:
- Implement blindfolded strategies
- Not sure if I want to compile Tom 7's C++ sources and get it to communicate the result with my program, or if I should redo it from scratch using python machine learning libraries...
- Implement NES Chessmaster communication
- Maybe also the chess engine from that weird cartridge I have for the original Game Boy?
- Add a bunch of the smaller strategies
- Implement a graphical chessboard, especially for human playing
- Make it so all games will output to a file to enable documenting games between the players
- Prolly more