... as EGC 2008 is coming close to the end. 8 rounds out of 10 in the main tournament has been played. And not so surprising Catalin Taranu from Romania is the number one European player at number 6. Top Swedish player is Girlwing at number 20.
Yesterday computer go tournament did not end so well for GNU Go. I wonder why. I have not looked at the games. Maybe somebody will put some regression test from the games.
At the moment I very much feel like programming on some go program. I actually would like to get my own program up to speed. But it would also be interesting to get some good things into GNU Go. I'm on the mailing list and what I see the activity is low. But the activity list for next minor release, 3.8, is long.
10 years ago
3 comments:
Gunnar Farnebäck held a talk on the advancement of go computer theory on wednesday. During that talk he mentioned (of course) GnuGo quite a bit.
His recomendation was that if you would like to contribute it would probably be best to focus on the montecarlo parts (wich apparantly is a small-ish subset of GnuGo).
Everybody do Monte Carlo now. It's because it scales well with parallel processing. But it does not solve the problem.
GNU Go 3.8 is out now. I played a game against it on 9x9 using default settings with monte-carlo and it played really bad. Much worse than with the standard engine. Did you check out the code? I am a bit tempted to do so and see how bad I can make it. :)
Post a Comment