PyTuroChamp

A family of toy chess engines inspired by Alan Turing's 1948 TUROCHAMP

PyTuroChamp is closest to the chess engine in Turing's paper, but adds piece-square tables that can be tuned with the PSTAB parameter. A higher parameter means more aggressive forward movement. With PSTAB = 0,

1. e3

is favored like Turing's algorithm would.

Bare removes the Turing heuristics and quiescence search and only contains the bare minimum a chess engine needs to play: alpha-beta search and a piece-square table.

Newt also ditches the old heuristics and adds newer techniques like PV-based iterative deepening and an opening book. It is comparatively strong and fast.

PTC-Host lets you easily host games between the engines directly from Python, without the need for a chess GUI.

Pyturochamp.py does not actually reproduce the results of either the Turing paper or the Chessbase implementation for Fritz. But then again Turing's paper was meant as a proof-of-concept, so maybe reproducibility is not important.

Running the engines from a chess GUI

First, install the python-chess chess library: pip install python-chess

The recommended option on Linux or MacOS is to modify and use the included shell scripts (ptc, bare, newt).

It is also possible and perhaps easier—especially on Windows—to launch Python directly from the GUI as in the Arena screenshot below. (Note that no log or PGN files will be created then, because the working directory will be somewhere where Python cannot create files.)

If you want to use one of the other engines besides pyturochamp.py, add "bare" or "newt" as additional command line parameters.

Files

Filename	Description
pyturochamp.py	The chess engine with Turing's heuristics. Plays more human-like, except for weird but typical moves like a2a4 and h2h4.
bare.py	Bare bones version, only alpha-beta and piece-square tables are used. Very computer-like and not pretty but sometimes efficient play. Stockfish took 62 moves to checkmate it (with ponder off).
newt.pt	Like Bare, this one ditches the heuristics. It adds principal variation (PV)-based iterative deepening and quiescence search like PyTuroChamp and also an opening book.
ptc, bare, newt	Shells script to run PTC/Bare/Newt from a chess GUI, e.g. KDE Knights (works very well) or XBoard. (Change the directory path inside first.)
ptc-host.py	Hosts a game between PyTuroChamp as White and Bare as Black. Updated board images are written to board.svg. (During play, board.svg should be opened in an image viewer that automatically reloads changed files.)
ptc_xboard.py	Combined XBoard and UCI interface module for PTC/Bare/Newt. Moves will also be logged to a PGN file.
movetest.py	Test engine responses to board situations
pst.py	Helper file with piece-square tables

In the icons directory, there are several logos in BMP format for various chess engine GUIs which were contributed by Norbert Raimund Leisner.

Improving performance by using PyPy

Running the scripts with PyPy3 instead of python3 will make the engines run about twice as fast.

Below is a sample terminal session that shows how to set up PyPy under Arch Linux and run the PyTuroChamp scripts. (Note that the "--local" command line switch is used here to install pip and python-chess into .local/bin/ in the user's home directory. This is optional, but perhaps a good idea on Linux. It also means that root permissions are not necessary.)

$ sudo pacman -S pypy3

$ pypy3 -m ensurepip --user

$ ls -l .local/bin
insgesamt 12
-rwxr-xr-x 1 martin users 232  3. Dez 20:33 easy_install-3.5
-rwxr-xr-x 1 martin users 204  3. Dez 20:33 pip3
-rwxr-xr-x 1 martin users 204  3. Dez 20:33 pip3.5

$ .local/bin/pip3 install python-chess --user

$ .local/bin/pip3 list
cffi (1.11.1)
greenlet (0.4.12)
pip (9.0.1)
python-chess (0.22.0)
readline (6.2.4.1)
setuptools (28.8.0)

$ pypy3 ptc_xboard.py newt
go
#    ()
move g2g3
quit

$ pypy3 pyturochamp.py
r n b q k b n r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
R N B Q K B N R
0.0
Your move? e2e4
r n b q k b n r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . P . . .
. . . . . . . .
P P P P . P P P
R N B Q K B N R
0.0
FEN: rnbqkbnr/pppppppp/8/8/4P3/8/PPPP1PPP/RNBQKBNR b KQkq - 0 1
(1/20) g8h6 -2.2 0.00
(2/20) g8f6 -3.9 0.00
(3/20) b8c6 -3.6 0.00
(4/20) b8a6 -2.0 0.00
(5/20) h7h6 -2.4 0.00
(6/20) g7g6 -2.3 0.00
(7/20) f7f6 -0.7 0.00
(8/20) e7e6 -6.7 0.00
(9/20) d7d6 -5.0 0.00
(10/20) c7c6 -3.1 0.00
(11/20) b7b6 -2.3 0.00
(12/20) a7a6 -2.2 0.00
(13/20) h7h5 -3.2 0.00
(14/20) g7g5 -2.4 0.00
(15/20) f7f5 -1.2 1.00
(16/20) e7e5 -7.3 0.00
(17/20) d7d5 -6.6 0.00
(18/20) c7c5 -3.6 0.00
(19/20) b7b5 -2.2 1.00
(20/20) a7a5 -3.0 0.00
# -7.30 ['e7e5']
My move: 1. e7e5     ( calculation time spent: 0 m 6 s )
r n b q k b n r
p p p p . p p p
. . . . . . . .
. . . . p . . .
. . . . P . . .
. . . . . . . .
P P P P . P P P
R N B Q K B N R
0.0
Your move?

Prerequisites

• Python 3 is recommended, but Python 2 should also work
• python-chess

References

• Turing, Alan (1952): Digital computers applied to games
• Chess Programming Wiki
• Muller, H.G.: Micro-Max, a 133-line Chess Source

License

• Public Domain
• The opening book and the piece-square tables from Sunfish are licensed under the GPL.