On 12/30/2010 08:20 PM, Aja wrote:
Hi Jeff,
When, do you think, did Mogo "started dominating all the KGS computer
events and CGOS, and also was the first to extend that dominance from
9x9 to 19x19."?
Hello Aja,
Here I quote from the computer-go archives, unless otherwise noted:
Dec. 31, 2006: John Tromp: "I spent most of yesterday on KGS playtesting
MoGo on 9x9 with 30 min total thinking time. The experience was quite
unlike any other program I've played on 9x9 in the
past. [..] I feel that the shodan level go 9x9 programs have arrived..."
Jan 12, 2007: Don Dailey: "Someone needs to get their bot on CGOS and
end Mogo's reign of terror. A version of MoGo has achieved a CGOS rating
of well over 2300!"
Mar 04, 2007: Peter Drake: "Congratulations to MoGo on winning the KGS
tournament held earlier today: [..] Even under borderline "blitz"
conditions (18 minutes sudden death for 19x19), MoGo managed to beat
conventional programs like GNU Go. [..] Of course, MoGo also beat all
the other MC/UCT programs. How did MoGo do it?"
Nick Wedd's summary from the above KGS tournament: "Last September
Kocsis and Csaba Szepesvári published Bandit based Monte-Carlo Planning.
The algorithm described there was implemented in MoGo. There was some
doubt about its effectiveness for large boards and for fast games, but
it was clear that it worked well for small boards particularly with slow
time limits: MoGo has won nine of the eleven KGS bot tournaments held
since September. However, this March event used full-sized boards and
fast time limits. Some people expected MoGo to do less well, perhaps
losing to more conventional programs such as ManyFaces, GNU Go, and Aya.
They were wrong. MoGo won all twelve of its games." [CrazyStone won 6
games in the same tournament.]
http://www.weddslist.com/kgs/past/24/index.html
Mar 17, 2007: Don Dailey: "It's unbelievable how strong MoGo is playing.
I remember when CGOS first came up, I expected it to be a
few years before a program could achieve 2000.0 on the CGOS
scale. But I was quickly surpised when programs started
breaking over 1800.0.
But this is quite incredible. MoGo_G3.4 at 2480! If you
look at the crosstable, it's mostly 100% wins against
everyone else. In fact, the only non-Mogo program to
beat it, won a single game out of 15 played, and it's
a pretty strong program too."
In Computer Olympiad 2007, Steenvreter was gold medal on 9x9.
Yes, and this is what Eric had to say about it: "Steenvreter uses UCT
and has some L&D knowledge that I reused from Magog."
and: "Steenvreter was really a rush job, hacking things together until
the last day before the tournament and no time to test properly. I was
hoping to be able to catch up with the stronger programs, but never
expected it to win the tournament."
Obviously it was following MoGo's lead with UCT (the tournament was held
in June 2007, well after the remarkable success of MoGo). I don't mean
to discredit Steenvreter, CrazyStone, or any other program. I'm just
focusing on MoGo so much because it set the bar so high and got
everybody chasing it.
Mogo's biggest contributions, so far, in my view, are
1.Applied UCT to computer Go, and such application came from the idea
"MCTS" that proposed in 2006 by Remi Coulom. Crazy Stone was using MCTS
to win 9x9 in 2006 Computer Olympiad.
2.See 3x3 patterns around the previous move.
3.RAVE (strictly speaking, it is invented by David Silver).
This is essentially what I said in my first message. I did not place
enough emphasis on CrazyStone then, even though I did reference it ("The
MoGo team applied UCT to Go with great success, using the idea of
building incremental trees from CrazyStone.").
UCT and RAVE are for both for the tree search. I think Crazy tone's
contribution for the playout is of same/or more important, because the
quality of simulations decide the playing strength much. From this view,
we should give Crazy Stone more and more credit.
Sure, I agree. CrazyStone was definitely a big part of the monte carlo
tree search revolution in computer go.
_______________________________________________
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
