gosu
a freeware chess program
current version: 0.16
Gosu is a freeware chess playing program. I hope you will enjoy it.
To play with it, you need to download a graphical interface:
WinBoard
or Arena.
"Gosu" is a Korean word that means "an expert, someone who has high skill at something".
Downloads:
Windows version:
gosu-0.16.zip - size 1678KB
Linux version:
gosu-0.16.tgz - size 2040KB
Command line options:
log=false - turns off the logging
ht_size=X - changes hash table size to X MB (you can pick a fractional number, e.g. ht_size=61.5)
Note: Gosu uses ~67MB of RAM with the default settings (ht_size=64)
Strength:
Estimated playing strength: 2300-2400 ELO.
Gosu plays in many tournaments and leagues. Here are some results:
| Event |
Place |
Participants |
Notes |
Version |
Date |
|
MPPS-5 |
3 |
12 |
Championship of Poland |
0.15 |
16-17 IX 2006 |
|
MPPS-4 |
1 |
8 |
Championship of Poland |
0.11 |
18-19 VI 2005 |
|
CCT7 |
12 |
44 |
tournament on the ICC |
0.6 |
12-13 II 2005 |
|
MPPS-3 |
5 |
9 |
Championship of Poland |
0.1/0.2 |
16-19 IX 2004 |
Long Internet events:
Thanks to all tournaments organizers for your great work.
Technical details:
Gosu uses two supervised learning methods for choosing the parameters of the evaluation function.
Both methods are based on GLEM by Michael Buro.
Some details are described in my M.Sc. thesis (in Polish only):
"Modelowanie funkcji oceniajacej w szachach"
(that means "Evaluation function modelling in chess").
Besides that Gosu uses many standard techniques like:
0x88 board representation, MTD(f), checkless qsearch, static exchange
evaluation, popular extensions
(check, 1-reply, pawn push, mate threat), lazy evaluation, opening book, etc.
History:
v.0.1 - 15 IX 2004:
Initial release, it played the first two games in MPPS-3
v.0.2 - 19 IX 2004,
I made a few changes during MPPS-3:
- fixed bug in pondering,
- ad hoc solution to mating problems,
- better recognizing of trapped bishop situations,
- better time control.
v.0.3 - 22 IX 2004:
- fixed bug causing storing illegal moves in the hash table,
- changed default hash table size to 64 MB,
- decreased log size,
- added possibility to turn off the logging.
v.0.4 - 2 X 2004:
- fixed bug that caused occasional wrong draw claim
v.0.5 - 15 XI 2004:
- many small changes and bugfixes
v.0.6 - 19 I 2005:
This version took part in the CCT7
and took a good 12th place.
- added opening book
- changed time control
v.0.7 - 10 IV 2005:
- new evaluation function
v.0.8 - 27 IV 2005:
- bugfixes and improvements in the evaluation function
- all evaluation weights are automatically selected, except a few, which were modified by hand
v.0.9 - 28 V 2005:
- fixed bug with clearing castle flags
- some hand tuning of weights - increased material value
v.0.10 - 11 VI 2005:
- large tuning of evaluation weights
- changed parameters of lazy evaluation
v.0.11 - 21 VI 2005:
This version has won the 4th Polish Computer Chess Championship.
- minor tuning of evaluation weights
- new opening book
v.0.12 - 4 VIII 2006:
- changed the compiler (update 16 VIII - compiled again the Windows version)
- new opening book
v.0.13 - 1 IX 2006:
- a few minor fixes and improvements
v.0.14 - 8 IX 2006:
- changes in the evaluation function
v.0.15 - 17 IX 2006:
- increased material value
- new opening book
v.0.16 - 7 X 2006:
- major refactoring
Acknowledgements:
- to the Polish computer chess programmers community,
progszach,
for competition, motivation and for sharing their knowledge.
Thanks to Grzegorz Sidorowicz (the author of Armageddon), Marek Kolacz
(Butcher), Tomasz Michniewski (Tytan), Marek Strejczek (Nesik), Maciej
Pestka (Matacz), Piotr Cichy (Nanoszachy), Mateusz Luksik (Atak),
Mariusz Rostek (Sierzant), Antoni Szczepanski (Matant) and others,
- to dr Maciej Szmit and Piotr Dachtera, the organizers of Polish
Computer Chess Championships, an important motivation for me to develop
the program,
- to Tomasz Jadachowski for coming up with the name "gosu",
- to the authors of the open source programs: Prof. Robert Hyatt for Crafty, Fabien
Letouzey for Fruit, Tord Romstad for Glaurung, Tom Kerrigan for TSCP
and Jon Dart for Arasan,
- to Dr. Michael Buro, Dr. Andreas Nowatzyk and Siddhartha Chinthapally, for
their research on automatic tuning of evaluation function parameters,
- to Jakub Pawlewicz, Marek Biskup and Piotr Milos for insightful
advices.
Author:
I am a PhD student at the Institute of Informatics (Algorithms and Complexity Group)
at Warsaw University.
mailto: paterek (at) mimuw.edu.pl
Copyright (C) Arkadiusz Paterek, 2003-2006. All rights reserved.