[computer-go] thoughts on 100,000 cgos games : Test/Benchmark

[alain Baeckeroot]

Le Mardi 2 Mai 2006 02:33, David G Doshay a écrit :
> Reaching 100,000 games seems like a good milestone for reflection, so  
> I hope that others will share their thoughts.
> 
> Of course, it is a great thing that Don has done to provide this  
> service, and I thank him for his initiative and efforts.
> 
> I have more of a list of random observations than coherent opinions:
> 

5/ Looks trivial but : compared to "static" testing (cgtc-2.0, STS, GNU Go...)
playing lots of games gives a very trustable result of the average strengh.

As a GNU Go humble beginner contributor, i consider with suspicion the gnugo
regression suite because it is based mainly (only?) on its own games, and thus
follow only its own logic which is sometimes wrong. (even if passing more
tests has so far proven that the engine is stronger)

So here is my wish: a cgos 9x9 reg_genmove test suite, maybe split by strengh:
 Something like put in common the tests that each one has done for
his own bot, based on cgos mistakes_to_avoid/good_move_to_keep.

Anybody interested ?
Alain