README

Matilda - Go/Igo/Wéiqí/Baduk playing software

Matilda is a competitive computer Go playing engine and accompanying software.
Go is an ancient and beautiful strategy board game; you can read more about it
at http://senseis.xmp.net/?WhatIsGo.

Implementation-wise Matilda is a MCTS Mogo-like program. It is aimed at 64 bit
computers in shared memory, playing with Chinese rules via the Go Text Protocol.
It is versatile and optimized for speed in a lot of areas, though some changes
require a recompilation.

The relative strength of Matilda can be seen from playing on the CGOS (see
http://www.yss-aya.com/cgos/9x9/standings.html). It is currently much stronger
in smaller boards than in larger ones.


System Requirements
  - Linux, BSD, macOS or other POSIX 2004 compliant system
  - C99 compiler suite with support for OpenMP 3.0 (like GCC or clang)

Before using read the INSTALL file carefully, and at least modify the file
src/inc/config.h to your taste.


How to

You can play with Matilda out of the box using a text interface. For a graphical
interface you can connect it with any GTP-speaking program that supports Chinese
rules, like GoGui, available at http://gogui.sourceforge.net/.
Matilda also includes matilda-twogtp for self-play and benchmarking.


Copyright

All parts of Matilda are licensed as permissive free software, as described in
the file LICENSE that should accompany this document, except for the following
files. src/crc32.c, which was derived from another file, is distributed with the
same license as the original (public domain). The files contained in the
src/data/ directory may also be based on game records or other foreign files,
and may be in dubious or case-by-case specific licensing circumstances. For
legal enquiries contact the author of this software.

This project started as the practical component of a dissertation for the
obtention of a Masters Degree on Computer Science and Computer Engineering, from
the High Institute of Engineering of Lisbon, titled "Guiding Monte Carlo tree
searches with neural networks in the game of Go" (2016) by Gonçalo Mendes
Ferreira.