Zappa Xtreme does the impossible..

[Tony Thomas]

It scored the same versus the same opponents in my tournaments. These sort of things happen very rarely, it goes to show that there isnt much difference between new version and the old one. Both of them scored 58/128 against the common opponents.

Code: Select all

 27 Gandalf 6.01                   : 2691   43  43   200    48.8 %   2700   22.5 %
 28 Scorpio 2.0                    : 2691   55  55   132    48.5 %   2701   15.2 %
 29 Scorpio 1.84 JA                : 2676   44  44   185    45.9 %   2704   23.8 %
 30 Delfi 5.0                      : 2672   54  54   119    47.1 %   2693   26.9 %
 31 Zappa_mexico                   : 2670   49  49   156    45.8 %   2699   21.2 %
 32 Pharaon 3.5.1                  : 2670   41  41   208    45.4 %   2702   24.5 %
 33 Zappa_mexico fix               : 2670   55  55   128    45.3 %   2702   18.8 %
 34 DeepSjeng25                    : 2666   49  50   148    45.9 %   2694   23.0 %
 35 Glaurung 1.2.1                 : 2664   50  50   143    44.8 %   2700   23.8 %
 36 WildCat 7.0                    : 2657   45  45   200    43.8 %   2700   14.5 %
 37 Ruffian 1.0.5                  : 2655   45  45   200    43.5 %   2700   15.0 %
 38 Prodeo 1.2                     : 2655   41  41   200    43.5 %   2700   28.0 %
 39 Crafty 21.6 JA                 : 2652   53  53   132    42.8 %   2702   22.0 %

Code: Select all

33 Zappa_mexico fix          : 2670  128 (+ 46,= 24,- 58), 45.3 %

Rybka v1.0 Beta.w32           :   4 (+  0,=  1,-  3), 12.5 %
WildCat 7.0                   :   4 (+  1,=  1,-  2), 37.5 %
TogaII 1.2 beta 2a KS/EHP     :   4 (+  0,=  2,-  2), 25.0 %
Spike 1.2 Turin               :   4 (+  2,=  0,-  2), 50.0 %
Smarthink 1.00                :   4 (+  1,=  1,-  2), 37.5 %
Prodeo 1.2                    :   4 (+  2,=  1,-  1), 62.5 %
Trace 1.37a                   :   4 (+  1,=  1,-  2), 37.5 %
Frenzee 3.0                   :   4 (+  3,=  0,-  1), 75.0 %
Gandalf 6.01                  :   4 (+  1,=  0,-  3), 25.0 %
Ktulu 8.0                     :   4 (+  3,=  1,-  0), 87.5 %
Thinker 4.7a                  :   4 (+  0,=  1,-  3), 12.5 %
Pharaon 3.5.1                 :   4 (+  1,=  0,-  3), 25.0 %
SOS 5.1                       :   4 (+  3,=  0,-  1), 75.0 %
Ruffian 1.0.5                 :   4 (+  3,=  0,-  1), 75.0 %
SlowChess Blitz WV 2.1        :   4 (+  1,=  1,-  2), 37.5 %
Aristarch 4.50                :   4 (+  3,=  1,-  0), 87.5 %
CM10th D2Alos                 :   4 (+  2,=  1,-  1), 62.5 %
HiarcsX50UCI                  :   4 (+  1,=  0,-  3), 25.0 %
List 512                      :   4 (+  0,=  2,-  2), 25.0 %
ChessTiger2007.1 UCI          :   4 (+  0,=  0,-  4),  0.0 %
Glaurung 2 Epsilon/5          :   4 (+  1,=  1,-  2), 37.5 %
Fruit 2.3                     :   4 (+  1,=  0,-  3), 25.0 %
Naum 2.2                      :   4 (+  0,=  0,-  4),  0.0 %
DeepSjeng27                   :   4 (+  3,=  1,-  0), 87.5 %
Zappa_mexico                  :   4 (+  0,=  3,-  1), 37.5 %
Delfi 5.2                     :   4 (+  3,=  0,-  1), 75.0 %
Movei00_8_438                 :   4 (+  3,=  1,-  0), 87.5 %
BugChess2_V1_5_2              :   4 (+  1,=  1,-  2), 37.5 %
Shredder11UCI                 :   4 (+  1,=  1,-  2), 37.5 %
Crafty 21.6 JA                :   4 (+  2,=  2,-  0), 75.0 %
Scorpio 2.0                   :   4 (+  2,=  0,-  2), 50.0 %
AlaricWB707                   :   4 (+  1,=  0,-  3), 25.0 %

[Uri Blass]

Somethings seem to be wrong with your results

Even engines that are weaker than movei like Delfi5.0 or Gandalf6.01 have better rating than Zappa.

For comparison the ccrl 40/4 has the following data



Zappa Mexico 32-bit 2875 +50 −49 57.4% −51.8 43.0% 128
52.3%
Zap!Chess Zanzibar 32-bit 2850 +37 −37 52.5% −17.5 35.8% 240
54.0%

Gandalf 6 2703 +24 −24 37.9% +86.6 30.5% 619
50.6%
Delfi 5.0 2677 +13 −13 35.4% +111.1 28.4% 2246

I wonder if zappa use some book in your test or play without book.

If zappa does not use book it may help to explain zappa's bad results
in your tests.

Uri

[swami]

I guess Anthony said it upteenth time that Zappa doesn't play at its current level in blitz time controls and it plays well at longer time controls and faster hardware.He then added that depth 10 and depth 15 are totally different animal.

I think that definition of depth difference only applies to engines that play stronger at longer time controls,but unbelievably weaker at faster time controls.Other Engines that belong to this category are Homer,Tytan etc

Not many engines exhibit this type of problem, I know only few.

[Tony Thomas]

Somethings seem to be wrong with your results

Even engines that are weaker than movei like Delfi5.0 or Gandalf6.01 have better rating than Zappa.

For comparison the ccrl 40/4 has the following data



Zappa Mexico 32-bit 2875 +50 −49 57.4% −51.8 43.0% 128
52.3%
Zap!Chess Zanzibar 32-bit 2850 +37 −37 52.5% −17.5 35.8% 240
54.0%

Gandalf 6 2703 +24 −24 37.9% +86.6 30.5% 619
50.6%
Delfi 5.0 2677 +13 −13 35.4% +111.1 28.4% 2246

I wonder if zappa use some book in your test or play without book.

If zappa does not use book it may help to explain zappa's bad results
in your tests.

Uri

Zappa just doesnt play well in my time control. Yes, Zappa does use a book. First version used Rybka 2.0 abk by wael deeb. Second version (the fix) used the book from the last public release, 1.1.

[Tony Thomas]

I guess Anthony said it upteenth time that Zappa doesn't play at its current level in blitz time controls and it plays well at longer time controls and faster hardware.He then added that depth 10 and depth 15 are totally different animal.

I think that definition of depth difference only applies to engines that play stronger at longer time controls,but unbelievably weaker at faster time controls.Other Engines that belong to this category are Homer,Tytan etc

Not many engines exhibit this type of problem, I know only few.

Ant also comes to mind.

[Uri Blass]

I guess Anthony said it upteenth time that Zappa doesn't play at its current level in blitz time controls and it plays well at longer time controls and faster hardware.He then added that depth 10 and depth 15 are totally different animal.

I think that definition of depth difference only applies to engines that play stronger at longer time controls,but unbelievably weaker at faster time controls.Other Engines that belong to this category are Homer,Tytan etc

Not many engines exhibit this type of problem, I know only few.

I compared with the ccrl 40/4 list and not with the ccrl 40/40 list.

Another point is that I do not see that zappa earns from long time control relative to delfi5.0 that is in Tony Thomas list and the difference seem to be almost 200 elo at all time controls.

From the 40/40 rating list

http://computerchess.org.uk/ccrl/4040.l ... t_all.html

Zappa Mexico 32-bit 2889 +22 −22 55.6% −35.2 43.5% 649
Zap!Chess Zanzibar 32-bit 2884 +22 −22 57.1% −47.5 43.6% 645
51
Delfi 5.0 2702 +20 −20 47.5% +15.4 36.8% 774

rating difference 182-187 elo

From the 40/4 rating list

http://computerchess.org.uk/ccrl/404/ra ... t_all.html

Zappa Mexico 32-bit 2875 +50 −49 57.4% −51.8 43.0% 128
Zap!Chess Zanzibar 32-bit 2850 +37 −37 52.5% −17.5 35.8% 240
Delfi 5.0 2677 +13 −13 35.4% +111.1 28.4% 2246

rating difference 173-198 elo

[Tony Thomas]

It is possible that Zappa just doesnt like incremental time controls, or my computer.

[GenoM]

yes, its interesting problem -- some engines do not like incremental time controls and make it much better on Xmoves/Yminutes time controls.
Time managing is dark area, I remember that I pointed out this 2 or 3 years ago: that engines are still very very faraway from good, human time management /more thinking in complicated positions/. Now Ossipov is making experiments with time management of Strelka. Only with time management improvement Strelka has /minimum/ +50 ELO by my estimates.

[Tony Thomas]

Are you saying that you have insider info on Strelka? Can you get me in the team as well?

[dr_zied_haddad]

Hi
I'm interested to become a beta tester of some Free UCI engines or also of any of the commercials.
I have many interesting ideas to improve the evaluation of engines and also special testing procedures as i did before when i discovered an excellent setting for Shredder engine which improved Shredder's rating in fast time controls by more than 60 rating points.

My mail: dr_zied_haddad(@)yahoo.fr