5p versus computer player on KGS

[wastelands]

I'm surprised this wasn't already posted. :-) I missed the live 9x9 games myself, but took a look at them using IaGoChall's games list. As far as I can tell, the human won two games, and the bot won one. 19x19 is still yet to be played, in about 13.5 hours. Here's the info from IaGoChall's User Data on KGS.

-----

In partnership with the Institut National de Recherche en Informatique et en Automatique (INRIA), a unique challenge will be held in parallel to the Paris Go tournament :

IA-GO Challenge 2008 - Challenge of Artificial Intelligence and Go

Developed by INRIA, Mogo ( http://www.inria.fr/saclay/ressources/zooms/mogo ), which is currently one of the best Go programs in the world, will challenge the professional Go player Catalin Taranu 5P. Mogo has all the computing power of INRIA with hundreds of super-computers in a network.

The winner will be appointed at the end of games played on a 9x9 board after 3 parts of 2x30-minute sudden death. A 19x19 exhibition will be held on the following day. This is the program of the event, to be followed live on KGS!

Saturday, March 22, 2008
3/22/08 8:00 AM Game I (9x9)
3/22/08 9:30 AM Game II 9x9
3/22/08 11:00 AM Game III 9x9

Sunday, March 23, 2008
3/22/08 8:00 AM Exhibition game (19x19)

More on http://paris2008.jeudego.org

[Harleqin]

Where can I find the games now? It seems that he didn't use his usual account.

[Wildclaw]

Originally Posted by Harleqin

Where can I find the games now? It seems that he didn't use his usual account.

http://www.gokgs.com/gameArchives.jsp?user=IaGoChall

[Tobiume]

I am eager to see what progress Mogo made compared to its last years performance against Guo Juan on 19x19.

[gauss]

Here's the 9x9 game that Catalin Taranu (B) lost to Mogo (W). Komi is 7.5. He seemed to crush Mogo in the other two games (taking W).

Anyone wish to comment on it? What was the losing move?


Download this SGF file.
(I've removed comments made at the end of the game; the sgf player here was having trouble with the quantity of them. They were all just chatter, I think.)

[Harleqin]

Well, in the end Black was trailing by about 2 points. I think that overall, 7.5 komi is too high for 9x9 -- my feeling is that 5 is about "correct".

[isshoni]

On IaGoChall's user profile, the record is
White - 4 victories
Black - none

Catalin played Mogo as white twice, Mogo played as white against Catalin once and against Bao Yun once.

[Wildclaw]

Originally Posted by Harleqin

Well, in the end Black was trailing by about 2 points.

Remember that Mogo usually wins by 0.5 points in the end. It has no interest in maximizing the points it wins with. If it is ahead it will play safer moves even if it loses a few points, so looking at the end result and judging how far ahead white was doesn't work very well.

Originally Posted by Harleqin

I think that overall, 7.5 komi is too high for 9x9 -- my feeling is that 5 is about "correct".

There is a bunch of evidence (mostly lots of computer games) that indicate that 7.5 komi (area scoring) is the closest fractional komi you can get on 9x9.

Originally Posted by gauss

Anyone wish to comment on it? What was the losing move?

Heh, that is a difficult question.

1. 5-5 is by far the most common opening move on 9x9. I am not saying that 3-4 is incorrect, but it does allow white a much stronger foothold in the center.

7. Allows white to get the c4 which proves lethal later on.

9. Is this the right invasion point? It does threaten to connect under, but after white blocks, black has to play c6 and the resulting shape isn't strong enough to help support the d3 stone. Still, I can't see any better ways to do it.

13. Black isn't strong enough on the top left to do this, I think. Still, what else is black supposed to do?

17. Can black do anything else here? Anyway, after this move white simply solidifies his position and black can do little as far as I can see.

Someone higher ranked could probably give a better analysis than me though.

[SpongeBob]

Does anyone know if Catalin is an expert also on the 9x9?

9x9 is very special and much different from 19x19. I think that if Catalin is not an expert on 9x9, his 19x19 strenght will help him, but it might not be enough against MoGo or against an amateur player with lots of experience on the smallest board. Maybe he joined this tournament mainly with regards to his game against MoGo on the 19x19. He did very well there with 9 handicap stones against him. Or maybe computers are really that strong already on the 9x9 ...

[breakfast]

I commented all 3 games of the match for Go4Go.
If someone wants to upload the commentaries on their sites, please contact me.
It's better, if your sites are computer-Go related.
I will ask only the link on your page in return.