Author: Uri Blass
Date: 15:26:50 12/10/00
Go up one level in this thread
On December 10, 2000 at 17:33:14, Thorsten Czub wrote:
<snipped>
>The draw-score moves: 29,30,31,32,33,34,35,37,38,39,40,41 shows exactly the area
>Gambit-tiger gets the evaluation increase.
>
>all the listed moves before, have nearly draw scores !
>and now look at gambit-tigers scores in the same time:
>this gambit-tiger score sees the plot.
>
>this "behaviour" is in fritz for many versions.
>it was world-championship paderborn and it was in other
>situations. i guess they cannot fix it, otherwise
>they would have done it much earlier.
I do not know if they can fix it or cannot fix it.
The fact that they did not fix it in the past does not prove that they cannot
fix it in the future.
It is possible that it takes them time to think about the right idea how to fix
it.
>it must be a major problem. if they would fix it,
>fritz would lose strength.
If they fix it in the wrong way they create other problems that are bigger.
I guess that it is not easy to fix it in the right way but it is not a proof
that it is impossible.
A possible way to try to fix it is to try to change the evaluation function if
the main line and the evaluation of fritz suggest that it has no idea about the
position to start using another engine.
I do not have Fritz6a but I see stupid lines of Fritz5.32
for example:29.Be2 Rc6 30.Bd3 at depth 10 with +0.22 pawns for white.
When Fritz suggests lines when white waste tempo and has an advantage then
it suggests that fritz has no idea about what happens so maybe it should change
the evaluation function.
>
>you can forget fritz in those situations. it is BLIND !
>guess a boxer, rocky, part V, suddenly gets blind
>from minute 29 to minute 41 !
>
>what would happen ? knocked out !
>and that is what gambit-tiger is doing with fritz.
>
>how do you want to use an "often blind" program for analysis ?
>how do you want to win a tournament or championship with a
>blind program?
>it makes 400.000 NPS or even more. say 1.000.000 NPS.
>and as a result you can replace all the Nodes by:
>
>evaluation:=0.00;
>
>
>no.
>fritz seems to see draw by repetitions for black !
>it believes: black is weaker, therefore it would like to draw.
>and the best i create some fresh draw lines.
>it computes 500.000 NPS to find out NOTHING about the position.
>
>that of course the reason there is no fritz7 on the market.
They simply do not call it Fritz7 but Deep Fritz.
Deep Fritz is supposed to be better than Fritz6 also in cases that you use one
processor.
I am interested to know if Deep fritz has also the draw evaluation in the game
that you posted.
>they can't bring it out. it would lose to gambit-tiger.
I remember that you said that it does 50% against gambittiger when gambittiger
is better only because it is better in beating other programs.
>thats the reason the latest computer-schach and spiele
>has nothing about fritz and much about shredder5.
>because fritz is losing against gambit-tiger by
>2xn to n (with n=number of winning points) against
>Gambit-Tiger.
Fritz6a is losing against Gambit but the question is if Deep Fritz is losing
against Gambittiger
>
>they HAVE no product. they had one, but when rebel was released
>they had none. so they now try to adapt.
>but how to fix something they have been unable to fix
>for many versions. the "bug" was in fritz6, in fritz5.32,
>it was in fritz4 and in fritz3.
>this shit draw-behaviour is part of fritz' strength.
>how to fix it ?!
Fritz3,4,5 and old Fritz6 were also preprocessors and I understood that they
fixed the problem in Fritz6a so maybe they can also fix other problems in the
future.
>
>they can only operate the beast and stand the opponent
>making jokes about their program and about the draw-score.
>making a bored face when opponent puts a finger on the draw-score.
>thats all they can do.
>
>
>[Event "Stufe=120'/40+60'/20+60'20. "]
>[Site "k6-400"]
>[Date "2000.12.08"]
>[Round "1"]
>[White "Fritz 6b"]
>[Black "Gambit-Tiger1.0"]
>[Result "0-1"]
>[PlyCount "86"]
>[EventDate "2000.12.08"]
>
>1. Nf3 c5 2. g3 Nc6 3. Bg2 g6 4. O-O Bg7 5. c4 Nf6 6. Nc3 O-O 7. d4 cxd4 8.
>Nxd4 Nxd4 9. Qxd4 d6 10. Qd3 a6 11. Bd2 Rb8 12. Rac1 Nd7 13. b3 Nc5 14. Qb1 b5
>15. cxb5 {0.25/14} 15... axb5 {+0.06} 16. b4 {0.16/14} 16... Na6 {-0.10} 17. e4
>{0.13/14} 17...
>Nc7 {+0.16} 18. Qd3 {0.25/12} 18... Qd7 {+0.14} 19. Rc2 {0.31/14} 19... Bb7
>{+0.14} 20. Rfc1 {0.31/14}
>20... Rfc8 {+0.20} 21. Be3 {0.25/13} 21... Ra8 {+0.22} 22. Bd4 {0.25/13} 22...
>Na6 {+0.36} 23. a3 {
>0.31/13} 23... Bxd4 {+0.32} 24. Qxd4 {0.25/13} 24... Nc7 {+0.16} 25. Bf1
>{0.34/13} 25... Ne6 {+0.18}
>26. Qb6 {0.25/14} 26... Rxa3 {+0.32} 27. Bxb5 {0.22/13} 27... Qc7 {+0.40} 28.
>Qe3 {0.16/13}
>28... Qd8 {+0.32} 29. Bd3 {-0.12/13} 29... h6 {0.42} 30. f4 {0.13/12} 30... Rb3
>{+0.80} 31. f5 {
>-0.06/13} 31... gxf5 {+0.74} 32. exf5 {-0.12/13} 32... Ng5 {+0.78} 33. h4
>{-0.19/13} 33... Nf3+ {+1.58 !!}
>34. Kf1 {0.00/13} 34... Ne5 {+1.38} 35. Be2 {0.00/12} 35... Qd7 {+1.80} 36. Qf4
>{-0.56/12}
>36... Ba8 {+1.82} 37. b5 {0.16/12} 37... Qb7 {+2.06 !!} 38. Ke1 {-0.12/14} 38...
>Kh7 {+2.56} 39. Qf2 {
>0.00/12} 39... Rc5 {+2.08} 40. Kd2 {0.06/12} 40... f6 {+2.40} 41. g4 {0.00/13}
>41... Rd5+ {+2.42} 42.
>Nxd5 {-1.16/15} 42... Qxd5+ {+3.98} 43. Ke1 {-1.84/16} 43... Rh3 {+4.08} 0-1
>
>How shall fritz6 win or stand games it shows NO IDEA about what happens on the
>board ?!
>
>In move 37. It even believes it is better again. In the moment is sees the
>problem (move 41.) it is too late and it loses very fast.
>
>The draw-score moves: 29,30,31,32,33,34,35,37,38,39,40,41 shows exactly the area
>Gambit-tiger gets the evaluation increase.
>In the moments Gambit-Tiger SEES and idea and evaluates it, Fritz sees only
>Silly draw-scores or scores nearly sero.
The interesting question is what are Deep Fritz's scores at moves 29-35,37-41.
Uri
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.