Author: Robert Hyatt
Date: 14:35:02 11/10/00
Go up one level in this thread
On November 10, 2000 at 16:26:43, Peter Berger wrote:
>I feel a little guilty about this post as Gambit Tiger 1.0 _really_ is a
>marvellous prog and usually smokes Crafty without much difficulties in most
>pretty games on equal hardware( sufficient data to prove beyond any doubt ) and
>I can't remember having so much fun with another chessprogram before .
>
>Yet this paradigm babble somehow annoys me a little , especially since I had to
>view the few posted games which proved nothing ( usually nice Tiger wins in
>complicated endgames where it _really_ rules ) .
>
>OK , so what are the theories ?
>
>Thorsten Czub/Whittington-alias : "Gambit Tiger wins its games because it
>follows a new "intelligent" paradigm diving into the "fog" in most mysterious
>ways !"
>
>Hyatt : " Tiger has a _very_ good search and so Gambit can make up for its
>speculative king safety eval by reaching high depths and sometimes even take
>advantage of it . High eval is mainly due to over-estimation of open files
>against opponent's king which is sometimes right , sometimes wrong ! "
>
>Dr Hyatt sometimes has a pretty dubious way of expressing himself IMHO but here
>I have an example that _really_ simply seems to be valid data supporting one of
>the both opinions ( something I have desperately missed before ..) :
>
>Game played on 2 PIII 500s , Time control : Game in 1 hour / 10 seconds
>increment per move :
>
>[Event "Computer chess game"]
>[Site "pete's compi(s)"]
>[Date "2000.11.10"]
>[Round "1"]
>[White "wcrafty17.14.exe"]
>[Black "Gambit Tiger 1.0"]
>[Result "1-0"]
>[TimeControl "3600+10"]
>
>1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Qb6 8.
>Nb3 Nbd7 9. Qf3 Be7 10. O-O-O Qc7 11. Bd3 b5 12. Rhe1 Bb7 13. a3 h6 14. Bh4
>O-O 15. Qh3 Rfc8 16. Kb1 Nb6 17. g4 d5 18. e5 Ne4 19. Bxe4 Bxh4 20. Qxh4
>dxe4 21. Qf2 Nc4 22. Nxe4 Bd5 23. Rd3 b4 24. axb4 a5 25. Nd6 Nxd6 26. exd6
>Qxd6 27. bxa5 Rab8 28. f5 Qc6 29. fxe6 Bxe6 30. h3 Bc4 31. Rc3 Qa4 32. Re4
>Rb4 33. Qf4 Qb5 34. Nd2 Rxb2+ 35. Kc1 Ra2 36. Nxc4 Ra1+ 37. Kd2 Qb1 38. Rb3
>Qd1+ 39. Kc3 Ra2 40. Qd2 Qa1+ 41. Kb4 Kh7 42. Qc3 Rb8+ 43. Kc5 Qg1+ 44. Qe3
>Rc8+ 45. Kb5 Qa1 46. Re8 Rxe8 47. Qxe8
>{Black resigns} 1-0
>
>Somewhere around move 21. ...Nc4 Gambit Tiger +2.50
>
>Gambit Tiger on the left :
>
>28. ...Qc6 +1.52 depth 11
>29. ...Bxe6 +0.86 depth 13
>30. ...Bc4 +1.38 depth 11
>31. ...Qa4 +0.92 depth 12
>32. ...Rb4 +0.76 depth 12
>33. ...Qb5 -0.10 depth 12 Crafty 34. Nd2 +3.78
>34. ...Rxb2+ +0.24 depth 12 Crafty 35. Kc1 +4.11
>35. ...Ra2 +1.28 depth 11 Crafty 36. Nxc4 +4.13
>36. ...Ra1+ +0.12 depth 11 Crafty 37. Kd2 +4.05
>37. ...Qb1 0.00 depth 11 Crafty 38. Rb3 +4.00
>38. ...Qd1+ +1.50 depth 10 Crafty 39. Kc3 +3.87
>39. ...Ra2 +0.70 depth 10 Crafty 40. Qd2 +4.22
>40. ...Qa1+ +0.70 depth 11 Crafty 41. Kb4 +4.49
>41. ...Kh7 -0.10 depth 10 Crafty 42. Qc3 +4.86
>
>42. ...Rb8 -0.07 depth 11 ( other engines might have resigned here ..)
>43. ...Qg1+ -0.20 depth 11 !!!!!!! Crafty 44. Qe3 +5.24
>
>And only now Gambit Tiger gets suspicious and gets a first fail-low to the
>negative side .
>The "new paradigm" obviously won't avoid letting the prog look a little silly
>from time to time :)
>
>After 47. Qxe8 I decided to resign for the Gambit Tiger .
>
>pete
That is the kind of thing I would worry about, if Crafty was doing it. Crafty
is presently not particularly fast. I am pretty sure that I could make it at
least 2x faster, if I was ready to cast a lot of things into stone, and _really_
muck up the readability/understandability of the source.
At present, I consider this a "research project" where it is more important to
follow the software principle "design for change" rather than to worry about
that 2x speed gain. As a result, on equal hardware, it is most definitely
going to be slower than the opponent... Even if its NPS is higher, because
NPS is not all that is going on.
One day, when parts of the evaluation begin to "settle" they will get re-written
in a much more efficient way. But so long as changes are interesting, I want to
"keep it simple" in terms of simply doing what I want in the most direct way
possible, not in the cleverist/fastest/most obtuse way.
And that incurs a penalty I don't mind, because I can pick up that factor of two
with the right hardware...
I'm amazed that I say +5, tiger says 0.0.. that looks ugly IMHO.
This page took 0.03 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.