Author: Peter Berger
Date: 14:53:40 06/03/02
Once upon a time in cold, foggy, wintery February I read a few posts on CCC. One
had been written by German chesscomputer journalist Michael Scheidl.
He had created his very own Chessmaster 8000 personality : CM_Brain ( I seem to
remember it was quite a common one with the usual higher selectivity and king
safety) and posted his testresults against Yace at CCC which looked quite
promising.
Mr Thomas Lagershausen countered with a response :
>Second, a cm-personality that doesn´t beat yace is from an absolutly beginner.
It was late at night - I definitely kind of disagreed and it sounded rude to my
ears , so something inside made me reply:
"Do you have a cm-personlity that will beat Yace by a 60% margin ( as I assume
there are lots of experts working in the CM field) ?
OK, for the sake of discussion I don't believe such a thing really exists.
Let's choose a time control, I know you love slow games. Is 60/10 Fischer time
control on an Athlon1333 enough ?
Yace shouldn't have the same hardware. Let it play on a PIII 1200 instead.
I bet no Chessmaster personality will get 60% against Yace under these
conditions. Bet accepted ?
If you post or email a personality supposed to do the job I will play the match
(let's make it best of 20 ?!, or 40 ?!) including complete logfiles of course.
Regards,
pete"
No response at first (those cowards ;) ) .. . I already thought I had got away
with it - and yes, I really knew from my own testing that some random strong
Chessmaster personality definitely wouldn't be able to deliver this job anyway.
Well - but one of the true knights of the ChessMaster personality business
thought it should be doable amd a worthy challenge . Mr Bill McGaugh, creator of
the mighty and well-known MGX-personalities , who has literally played thousand
of games to tweak the ChessMaster always a little more to really get the best
out of it, decided to take the challenge.
By accident and experiment one of his creations had been especially evil to Yace
in his own endless testseries .
To quote him:
"I have a personality that is 46 wins , 3 losses , and 23 draws (79.7%)versus
Yace
0.99.56 within the chessmaster gui..."
Now I was doomed to my own words for about a 100 hours of manual engine-matches
:=) . My feeling before the match started was: Yace should probably win this one
, but it might be close and Mr McGaugh seemed to feel just the same. And - yes,
already then I thought my wording had been _very_ unfortunate ( that Yace would
have to score 40%+ to win), believe it or not..
First I was enthusiastic and the first games were played quite fast.
To quote myself from email ( exact conditions for the "nerds" ;-) included)
"Hi Bill, hi Dieter,
time for a first brief report, this time still without the games and
logfiles , just a description of the basic setup and the first results so
that you know it is still going on and I haven't forgotten about it.
CMmg6 plays on an Athlon AMD 1.33 GHz,256MB RAM with 128 MB Hashtables and
uses the CM8000.opk and the Chessmaster Interface.
Yace 0.99.56 plays on my PIII1200 notebook, 512 MB RAM with 384MB hash, 8M
cache and the 3,4,5-Nalimov tablebases. It uses the medium annotated book
built by Mogens Larsen that I downloaded from its homepage and is playing
under WinBoard 4.2.6 .
Time control is game in 1 hour with 10 seconds increment per move. The match
is scheduled for 40 games with alternating colours, Yace had white in the
first game.
The challenge will be won for Bill McGaugh if CMmg6 reaches 24.0 points. It
will be won for Yace (and me ;-) ) if Yace reaches 16.5 points.
13 of the 40 games have been played, the 14th is scheduled for tonight .
After a slow and equal start mg6 increased the pace. Everything is still
possible of course but it seems mg6 has quite good chances to make it,
although Yace still has good chances, too. I already suspect the final match
result won't be significant in a scientific way at all.
Lots of the games were very interesting and I hope to be able to comment on
them later. If you prefer the raw data first, just drop me an email and I'll
send what I have so far.
Yace 0 1/2 1 1/2 0 1 0 0 1/2 0
0 1/2 1 5/13 ~38.5% still needs 11.5/27 to win the contest.
CMmg6 1 1/2 0 1/2 1 0 1 1 1/2 1 1
1/2 0 8/13 ~ 61.5% still needs 16.0/27 to win the contest.
Kind regards,
pete"
But it got tedious soon and also there are always so many other cool, thrilling
issues to test and do in computerchessland.
Yet - slowly the match continued .., and it hadn't been forgotten .
To cut a long story short - after 27 games Yace had managed to even equalize the
match score with 13.5/27 and my feeling was that this _definitely_ should be
enough to finish CMMG6 off, yet ..
While Yace only needed 3.5/13 then to finish the imposter off ;) , CMMG6 had an
incredible comeback to score 9.5/11 in the next games !!!!
The match remained close until the very last game and then it was:
Yace 0 1/2 1 1/2 0 1 0 0 1/2 0
0 1/2 1 1/2 1 1 1 0 0 0 0 1 1/2 1
1 1/2 1 0 0 0 1/2 1/2 0 1/2 0 0 0 0
1/2 1/2
16.0/40 =40.0%
CMmg6 1 1/2 0 1/2 1 0 1 1 1/2 1 1
1/2 0 1/2 0 0 0 1 1 1 1 0 1/2 0 0
0 1/2 0 1 1 1 1/2 1/2 1 1/2 1 1 1 1
1/2 1/2
24.0/40 =60.0%
Ouch !!
Congratulations, Mr Bill McGaugh - my prediction failed :-( ! By the closest of
margins at least .. I have lost my bet! You have won - your Chessmaster
personality in fact _has_ been able to score those 60% against Yace needed with
the ~15% hardware advantage :-( .
OK, no bad sports - but it was soo close ;) , grrr!
The scoresheet has quite some upsets IMHO - for example from rounds 12-17 Yace
scored 5/6 - fron round 22-27 Yace scored 5/6 again ... ; and yes: from rounds
31-38 the ChessMaster MG6 personality really scored an incredible 9.5/11 when it
really needed it badly enough !! So much about statistics ..
The match was thrilling to me and I had fun. But let's face it : we can't even
judge who is better with these results - as long as they took.
Many very interesting games and the prettiest have been won by the very winner
of the contest in fact. Yace won those technical games with its usual smooth,
correct style and as I like this precise,quiet style my love for it has in fact
grown further again - whenever the board really was on fire it was CMMg6 to
return as Phoenix though. I have never seen more instructive games in a
computers' match before that should make such nice testpositions in fact IMHO.
Only one brief testposition from round 35 and the shortest game of the match for
tonight. This is both a very interesting move to avoid IMHO and also a nice
example how mate attacks might be of importance _quite_ in advance to move 35 .
Yace 0.99.56,P - CM8000,A
[D]rn2kb1r/pQ5p/2p1pq2/8/P2PP3/2N1P2p/1P4PP/R1B3K1 w kq - 0 1 am Qxa8
taken from:
[Event "The Yace Challenge"]
[Site "PIII1200-Athlon1333,Ponder=On"]
[Date "2002.06.01"]
[Round "35"]
[White "Yace 0.99.56,PIII1200"]
[Black "CM8000,Athlon1333,MG6-Personality"]
[Result "0-1"]
[ECO "D17"]
[Opening "QGD Slav"]
[Variation "Czech defense"]
[Timecontrol "0 3600 +10 0 0"]
[Time "Sat Jun 01 20:50:42 2002"]
[last_book_move "11. O-O"]
1. Nf3 {0s} Nf6 {4s} 2. c4 {0s} c6 {3s} 3. d4 {0s} d5 {3s} 4. Nc3 {0s} dxc4
{3s} 5. a4 {0s} Bf5 {4s} 6. Nh4 {0s} Bc8 {4s} 7. e3 {0s} e5 {4s} 8. Bxc4 {0s}
e4 {91s} 9. f3 {0s} g5 {95s} 10. fxe4 {0s} gxh4 {46s} 11. O-O {0s} Be6 {64s}
12. Bxe6 {-0.27/12 324s} fxe6 {5s} 13. Qb3 {-0.15/12 234s} Qe7 {0s} 14. Rxf6
{-0.02/13 441s} Qxf6 {32s} 15. Qxb7 {-0.02/12 91s} h3 {0s} 16. Qxa8
{-0.72/11 205s} Bd6 {43s} 17. e5 {-M10/13 328s} Qg6 {5s} 18. Qxb8+
{-M08/13 62s} Bxb8 {5s} 19. Kf2 {-M07/14 31s} Qc2+ {4s} 20. Bd2 {-M06/12 0s}
0-1 {White resigns}
Cheers,
Peter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.