[ Halloween homepage | TOC | Best of | Usenet | Thanx | Steffens homepage | Steffens chess page ]

The Halloween-Attack in the Four Knight Game

---- Fritz5 vs. crafty match ----

(C)opyright 1996-1997 by Steffen A. Jakob

Re: ---- Fritz5 vs. crafty match ----

From           saj@elias.gams.co.at (Steffen A. Jakob)
Organization   g.a.m.s.
Date           18 Aug 1997 08:04:28 +0200
Newsgroups     rec.games.chess.computer
Message-ID     <u3u3gorpqr.fsf@elias.gams.co.at>

Hi Bob!

hyatt@crafty.cis.uab.edu (Robert Hyatt) writes:

> Steffen A. Jakob (saj@elias.gams.co.at) wrote:
> : I call this line the "Back-To-The-Roots"-line :-) Exchanging the queen
> : is surely not good for white. I have analyzed 10. Bf4 instead of
> : 10. Qe2+. I really would like to see Brause play this line with
> : Crafty. Can we arrange this? Can you force Crafty to play the same
> : with white till move #9?
> I can probably set up a special book with only that line if you want.  That
> would work.

As I posted before Brause already played with crafty.

> The most interesting thing I saw was that a "clean" crafty will play this
> differently than the "real" crafty, due to all the games you have had Brause
> play.  :)  The position learning let crafty walk a tightrope and win a game
> relatively easily even though it had no book to help it...

I wouldn't be so proud of what Crafty has learned so far. Did you
really have a look at the score of the last Brause vs. Crafty
game? Remember the first 14 moves:

[Event "ICS Rated Blitz match"]
[Site ""]
[Date "1997.08.14"]
[Round "-"]
[White "Brause"]
[Black "crafty"]
[Result "1-0"]
[WhiteElo "2370"]
[BlackElo "2484"]
[TimeControl "120+12"]

1. e4 e5 2. Nf3 Nf6 3. Nc3 Nc6 4. Nxe5 Nxe5 5. d4 Nc6 6. d5 Nb8 7. e5 Ng8
8. d6 cxd6 9. exd6 Qa5 10. Bf4 Qb4 11. Qd2 Qxb2 12. Rb1 Qa3 13. Nb5 Qxa2
14. Nc7+ Kd8


This is the result of Crafty's learning process after hundreds of
games in which it's opponent has sacrificed a piece in the 4th move.

- _All_ black rooks/bishops/knights/pawns are in their initial position
- the queen made 5 moves in a row and can now easily attacked
- what do you say about king safety? ;-)
- the white pawn d6 paralyses the black position
- white will get back material

Conclusion: this position is totally lost for black. Even if black
doesn't play 13... Qxb2?? white will get an advantage.

Bob, I think that Crafty's position learning feature is too slow. As I
already told you some months ago in a private mail, I am fascinated of
Nimzo's learning feature which is the most effective I have ever seen.

> First real case I've seen where the position learning stuff has made a
> real difference in the way it plays.  Warning to humans:  don't try this
> against Crafty.  :)

I really cannot agree if I have a look at Brause's Halloween
database. My advice: if you want to reach a +- position after few
moves against a strong computer (or human of course) then play the
Halloween Attack! ;-)

Best wishes,
Steffen A. Jakob          |"Die Ratten versenken das
saj@gams.at               | stinkende Schiff."
+43 1 8176230-18          | (Regina Jakob)
http://www.jakob.at/      |

© steffen ; last modification: $Date: 2005/11/16 22:18:35 $