regression/games/STS-RV/STS-RV_README

STS-RV 1.0

1. What is STS-RV?
2. How is it distributed?
3. Who created it?
4. How to use it?
5. Structure of the STS-RV
6. Acknowledgements


1. WHAT IS STS-RV?
-----------------
STS-RV is a Semeai Test Suite that uses GTP regression features with
722 semeai problems.


2. HOW IS IT DISTRIBUTED?
-------------------------
STS-RV is distributed under GPL license.


3. WHO CREATED IT?
------------------
STS-RV is part of the results of Ricard Vilà's master thesis. I worked
on my master thesis during six months at the AI research group in
Université Paris VIII under the direction of Tristan Cazenave and
Marco A. Peña.

To contact me: ritx@ya.com


4. HOW TO USE IT?
-----------------
If your program is compatible with GTP protocol you need to copy
the whole STS-RV contents in the folder where your executable program
is. The solve_semeaiS gtp command is not a standard one so probably
you will have to define your own gtp command for that purpose.
Then edit the file regression/evalSemeais.sh and change the name
of the solver to the one you use. Once this is done call the shell
script evalSemeaisAll.sh and all the tests will be tried. If you
create debug sgf files they will be redirected to the Debug folders
(change the name of the debug files in the evalSemeaisAll.sh to fit
your own names).

If your program is not compatible with GTP protocol check the official
GTP site at http://www.lysator.liu.se/~gunnar/gtp/

If you try the STS-RV i would like to hear from you, your opinions,
suggestions, found bugs, etc are of great interest to me so please
contact me at ritx@ya.com

5. STRUCTURE OF THE STS_RV
--------------------------
STS-RV is divided into six tst files containing a total of 130 sgf
files and 722 tests.
Every test is twice checked (playing black and white first).
Contents of the tst files:

semeais_0.tst containing semeai tests of class 0.
semeais_1.tst containing semeai tests of class 1 and 2.
semeais_e.tst containing semeai tests of class e.

All the tests in these files are created by me in order to check the
performance of what my module was supposed to do
correctly. Theoretically they are supposed to be passed statically (at
the root node).

If you are not familiar with this classification check Martin Müller's
paper: "Race to Capture: Analyzing Semeai in Go" at
http://www.cs.ualberta.ca/~mmueller/ or my master thesis "Desarrollo
del módulo de resolucion de semeai Semeai_01es" at
http://www.ai.univ-paris8.fr/~ritx/ (in Spanish).

semeais_RH.tst containing semeai tests obtained from "Counting
Liberties and Winning Capturing Races" written by Richard Hunter and
published by Slate and Shell (SSRH003). These are the most difficult
tests in the suite both for the number of possible moves at the root
node and for the depth of the winning sequence.

semeais_GSAT.tst containing semeai tests involving tesuji obtained
from "Get Strong at Tesuji" written by Richard Bozulich and published
by Kiseido Publishing Company (K56).

semeais_Misc.tst containing semeai tests over class 3 created by me
from my own games or compositions created by me.

6. ACKNOWLEDGEMENTS
-------------------
I would like to thank Martin Müller for settling the first stone of my
work with his interesting classification of semeais.

I thank Richard Hunter and Richard Bozulich for his express permission
to use the problems in their books as tests for machines, for the
benefit of the Computer Go community.

I would like to thank Gunnar Farnebäck for his valuable help
during the process of creating the suite.

I would also like to thank all the people in the AI research group at
the Université Paris VIII that have helped me to complete my work.

Finally i would like to thank the two researchers that co-directed my
work, Tristan Cazenave and Marco A. Peña.