What is the rank of Gnu go 3.8 (bot)

For discussing go computing, software announcements, etc.
Post Reply
Darsey
Dies with sente
Posts: 120
Joined: Tue Jun 17, 2014 4:35 pm
GD Posts: 0
KGS: 5k
IGS: 10K
OGS: 7k
Location: Salamanca, Spain
Has thanked: 37 times
Been thanked: 20 times

What is the rank of Gnu go 3.8 (bot)

Post by Darsey »

Good morning

I have in quarry (one program to play go for computer, I got it in my laptop with linux mint). I usually use it to practice some josekis, answers or see games. But I am man of risk and I want try be more strong that my computer, but the bot of quarry, that is GNU GO 3.8, always humbles me.

Therefore, I want know what rank is my archenemy GNU GO 3.8 when I want play without internet.
User avatar
pleiade67
Dies with sente
Posts: 71
Joined: Tue Mar 12, 2013 11:44 am
Rank: 20 kyu
GD Posts: 0
Has thanked: 10 times
Been thanked: 19 times

Re: What is the rank of Gnu go 3.8 (bot)

Post by pleiade67 »

The last stable versions of gnugo 3.7 and 3.8 (non Monte Carlo) are usually rated in the range 6k-5k KGS at their best.

Source:

http://senseis.xmp.net/?GNUGo
http://senseis.xmp.net/?KGSBotRatings
http://senseis.xmp.net/?KGSBots
Darsey
Dies with sente
Posts: 120
Joined: Tue Jun 17, 2014 4:35 pm
GD Posts: 0
KGS: 5k
IGS: 10K
OGS: 7k
Location: Salamanca, Spain
Has thanked: 37 times
Been thanked: 20 times

Re: What is the rank of Gnu go 3.8 (bot)

Post by Darsey »

Very thanks. I didn't know these articles in sensei.
Mike Novack
Lives in sente
Posts: 1045
Joined: Mon Aug 09, 2010 9:36 am
GD Posts: 0
Been thanked: 182 times

Re: What is the rank of Gnu go 3.8 (bot)

Post by Mike Novack »

If you need* to use the computer as an opponent don't try to play even games against a much stronger program or set the program's playing level weak enough to be able to play even games against it. You either will never see when you do things right vs doi them wrong (in the first case) or will learn the bad habits of that particular program/setting in the second.

Gnugo is using an AI (go knowledge based)based evaluator. It will play too passively, but OK for you to use until you learn how to beat it. Figure max strength at about 5-6 kyu (all a good AI go knowledge based evaluator can do).

Fuego uses a MCTS evaluator and will not play passively. When you get down to playing even games (or even 2 stone games) against gnugo switch to fuego. Meanwhile, begin taking handicap stones. First discover how many you need to win perhaps half of the games and then take a stone less. Ideally you want to lose 3-4 games for every one you win to maximize learning. The times you do win you presumably didn't make the same mistakes as you did in the games you lost and you might come to recognize the traps you avoided.






*for example I can't afford to tie up our phone line to play against a human on a go server, too long at a stretch.
Post Reply