Knockout tourney in progress

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 19 February 2004 00:06:52:

Settings
CPU: Dual P3-866MHz/384MB.
GUI: GalisWBTM and Winboard 4.2.7.
Hash: Max. 70MB + 16MB tbcache.
TB: 5-piece Nalimov.
Timecontrol:
Round 1+2: 4*G/60', Tiebreak: 2*G/30', 1*G/10'.
Semifinal: 4*G/90', Tiebreak: 2*G/45', 1*G/15'.
Final: 6*G/120', Tiebreak: 2*G/60', 1*G/20'.
Engines
Ruffian (1), Aristarch (2), El Chinito (3), Crafty (4), The King (5), Gandalf (6), Green Light Chess (7), Delfi (8), Dragon (9), SlowChess (10), Comet (11), Thinker (12), Ktulu (13), Postmodernist (14), Nejmet (15), Gothmog (16).
Ranking in parantheses. Placement of first four and last ten based on haphazard ranking/qualification tournaments. Adding The King and Gandalf inbetween.
Versions can be updated between rounds.

Round 1:
Ruffian 1.0.5 (1) - (16) Gothmog 0.4.5 : 2.5 - 0.5.
Aristarch 4.41 (2) - (15) Nejmet 3.07 : 3.0 - 0.0.
El Chinito 3.25 (3) - (14) PMv1009 : 2.5 - 0.5.
Crafty 19.10 (4) - (13) Ktulu 4.2 : 2.5 - 1.5.
The King 3.12d (5) - (12) Thinker 4.5a : 2.5 - 1.5.
Gandalf 5.1 (6) - (11) Comet B.68 : 1.5 - 2.5.
GLC v3.0 (7) - (10) SlowChess 2.89b: 2.5 - 0.5.
Delfi 4.4 (8) - (9) Dragon 4.5 : 2.5 - 1.5.
Round 2
Ruffian 1.0.5 (1) - (8) Delfi 4.4 :
Aristarch 4.41 (2) - (7) GLC v3.0 :
El Chinito 3.25 (3) - (11) Comet B.68 :
Crafty 19.10 (4) - (5) The King 3.12d :

Regards,
Mogens

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 19 February 2004 17:46:43:
Als Antwort auf: / In reply to: Re: Knockout tourney in progress geschrieben von: / posted by: Mogens Larsen at 19 February 2004 17:04:22:

Anyone playing the Evans gambit succesfully deserves to be in the next round.

My impression from watching my own engine play is that the Evans gambit
is actually a very effective opening for computers. White's compensation
seems to be of the kind that computers are rather good at understanding.
However, the number of games I have seen is not big enough to make any
certain conclusions. It is possible that my engine has just been lucky
in the few games I have seen.
In a few days or weeks, I plan to release a tournament book for Gothmog,
containing sharp lines like the Evans Gambit, the King's Gambit, and
the Dragon Sicilian. It will be interesting to see if it has any effect
on the results.
Tord

[Marcus Prewarski]

Geschrieben von: / Posted by: Marcus Prewarski at 19 February 2004 20:03:57:
Als Antwort auf: / In reply to: Re: Knockout tourney in progress geschrieben von: / posted by: Tord Romstad at 19 February 2004 19:30:29:

Anyone playing the Evans gambit succesfully deserves to be in the next round.

My impression from watching my own engine play is that the Evans gambit
is actually a very effective opening for computers. White's compensation
seems to be of the kind that computers are rather good at understanding.

My observations is similar. Same seems to hold for the Slavian gambit,
and perhaps even for the Benko gambit when playing black.

Which one is this?
This is most definitely not the case for my engine. Gothmog is really ugly
to watch on the black side of the Benko.
Another gambit which perhaps could work well for computers is the Blumenfeld
(1. d4 Nf6 2. c4 e6 3. Nf3 c5 4. d5 b5).
Tord

I agree with you about the Evans Gambit, it does seem well suited for computer play and the Kings Gambit also although to a lesser extent. I haven't seen my engine play many Sicilian Dragon games but I doubt its ability to play any flank opening well. In general I think my engine handles romantic era openings better, perhaps because there are fewer strategic ideas involved.
I've also had similiar experiences with the Benko gambit. Even though I like to play this it seems my engine has a hard time understanding it perhaps because the pressure is on the queen side of the board.
I have some other gambit ideas that seem promising.
DrunkenMaster still has a puny opening book of about 6000 positions. I started with a normal game collection but I think from here I'm going to shape it more by hand. One of the first things I did when I spent some time shaping it was take out 1. Nf3 , it seemed this was the best way for humans rated about 500 points lower on ICC to get draws. Next I may take out 1. c4 .
-Marcus

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 19 February 2004 23:19:57:
Als Antwort auf: / In reply to: Re: Knockout tourney in progress geschrieben von: / posted by: Tord Romstad at 19 February 2004 17:46:43:

My impression from watching my own engine play is that the Evans gambit
is actually a very effective opening for computers. White's compensation
seems to be of the kind that computers are rather good at understanding.
However, the number of games I have seen is not big enough to make any
certain conclusions. It is possible that my engine has just been lucky
in the few games I have seen.
In a few days or weeks, I plan to release a tournament book for Gothmog,
containing sharp lines like the Evans Gambit, the King's Gambit, and
the Dragon Sicilian. It will be interesting to see if it has any effect
on the results.

One advantage with offbeat opening variations is the reasonable probability of a not too sharp, theoretical reply from the opponent, which would disprove the soundness of the line. Sometimes it takes the opponent out of book and reinventing theory on the fly isn't always easy in sharp variations. Both scenarioes offers some kind of "compensation".
I intend to dabble a bit with the Benko gambit using Yace as sacrificial lamb. And a few offbeat Ruy Lopez variations like Berlin, Steinitz and maybe Bird. Not exactly sharp openings in general. But it's interesting to see what engines do when a GM names a certain line "unclear". Maybe I'll try to fiddle with the Benko for Gothmog when time permits. Implement some of the ideas by Steffen Pedersen ("The GAMBIT Guide to the Benko Gambit").
Regards,
Mogens

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 20 February 2004 07:44:24:
Als Antwort auf: / In reply to: Re: Knockout tourney in progress geschrieben von: / posted by: Mogens Larsen at 19 February 2004 23:19:57:

Maybe I'll try to fiddle with the Benko for Gothmog when time permits.

Please don't.
Gothmog doesn't understand the Benko at all. There is nothing Gothmog hates
more than playing positions where the opponent has a space advantage in the
centre. In the Benko, my engine typically thinks it is down by 1.5-2.0 pawns
at the time it is out of book. When it is so pessimistic about its own position,
it tends to self-destruct and lose without a fight.
I believe more in gambits like the Evans Gambit, the King's Gambit and the
Blumenfeld Gambit, where the attacker usually gets a nice pawn centre as
part of his compensation. Of course, I am only talking about how the openings
work for my engine here, and not about their objective soundness. I have no
reason to believe that the Benko is any less sound from a theoretical point
of view than the other gambits mentioned.
Tord

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 20 February 2004 17:02:46:
Als Antwort auf: / In reply to: Re: Knockout tourney in progress geschrieben von: / posted by: Tord Romstad at 20 February 2004 07:44:24:

Well, I might do it anyway (insert evil laughter). Adding the Blumenfeld as a response to Benko declined, ie. 1. d4 Nf6 2. c4 c5 3. d5 b5 4. Nf3 e6, looks like an interesting option.

Please don't.

- Hey! Don't do that. It isn't fair!
- Of course it isn't fair. I'm evil, look it up.

Regards,
Mogens