Komodo-Dragon-2 vs Stockfish 14 at knight odss

lkaufman · Post by **lkaufman** » Tue Sep 21, 2021 8:35 pm

Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2

Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29

Stockfish 14

Code: Select all

Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29

Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm

Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.

Rebel · Post by **Rebel** » Tue Sep 21, 2021 9:16 pm

lkaufman wrote: ↑Tue Sep 21, 2021 8:35 pm

Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2

Code: Select all

Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29

Stockfish 14

Code: Select all

Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29

Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm

Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.

I am not very familiar with contempt settings, stockfish indeed has no such setting but likely it has some default setting in the source code. How do we get them equal?

lkaufman · Post by **lkaufman** » Tue Sep 21, 2021 10:03 pm

Rebel wrote: ↑Tue Sep 21, 2021 9:16 pm

lkaufman wrote: ↑Tue Sep 21, 2021 8:35 pm

Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2

Code: Select all

Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29

Stockfish 14

Code: Select all

Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29

Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm

Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.

I am not very familiar with contempt settings, stockfish indeed has no such setting but likely it has some default setting in the source code. How do we get them equal?

Stockfish had a UCI option for Contempt but did away with it in SF14, though I don't know their reasons or whether they have a nonzero value set in the code. The Komodo Dragon value of 8 is much lower than what SF formerly used, but obviously higher than zero. If someone can verify that it does use zero in the code now, then you could lower the Dragon value from 8 to 0 to be equivalent, but that's a very minor change, hardly worth the bother. It might lower our winning percentage by one point or less, I would estimate.

Chessqueen · Post by **Chessqueen** » Tue Sep 21, 2021 11:05 pm

lkaufman wrote: ↑Tue Sep 21, 2021 10:03 pm
Rebel wrote: ↑Tue Sep 21, 2021 9:16 pm
lkaufman wrote: ↑Tue Sep 21, 2021 8:35 pm
Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2
Code: Select all
Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29
Stockfish 14
Code: Select all
Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29
Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm
Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.
I am not very familiar with contempt settings, stockfish indeed has no such setting but likely it has some default setting in the source code. How do we get them equal?
Stockfish had a UCI option for Contempt but did away with it in SF14, though I don't know their reasons or whether they have a nonzero value set in the code. The Komodo Dragon value of 8 is much lower than what SF formerly used, but obviously higher than zero. If someone can verify that it does use zero in the code now, then you could lower the Dragon value from 8 to 0 to be equivalent, but that's a very minor change, hardly worth the bother. It might lower our winning percentage by one point or less, I would estimate.

The contempt value of Stockfish 13 is set to 24 on the configuration, I do NOT know if it is set to 24 in the code, either if by setting it to 8 will help stockfish 13 to be as strong as Komodo Dragon2 giving KInight, and Rook Odds.

Uri Blass · Post by **Uri Blass** » Tue Sep 21, 2021 11:17 pm

Chessqueen wrote: ↑Tue Sep 21, 2021 11:05 pm
lkaufman wrote: ↑Tue Sep 21, 2021 10:03 pm
Rebel wrote: ↑Tue Sep 21, 2021 9:16 pm
lkaufman wrote: ↑Tue Sep 21, 2021 8:35 pm
Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2
Code: Select all
Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29
Stockfish 14
Code: Select all
Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29
Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm
Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.
I am not very familiar with contempt settings, stockfish indeed has no such setting but likely it has some default setting in the source code. How do we get them equal?
Stockfish had a UCI option for Contempt but did away with it in SF14, though I don't know their reasons or whether they have a nonzero value set in the code. The Komodo Dragon value of 8 is much lower than what SF formerly used, but obviously higher than zero. If someone can verify that it does use zero in the code now, then you could lower the Dragon value from 8 to 0 to be equivalent, but that's a very minor change, hardly worth the bother. It might lower our winning percentage by one point or less, I would estimate.
The contempt value of Stockfish 13 is set to 24 on the configuration, I do NOT know if it is set to 24 in the code, either if by setting it to 8 will help stockfish 13 to be as strong as Komodo Dragon2 giving KInight, and Rook Odds.

My guess is that Stockfish13 with the biggest possible contempt is going to get a clearly better result than Stockfish14.

lkaufman · Post by **lkaufman** » Wed Sep 22, 2021 12:08 am

Uri Blass wrote: ↑Tue Sep 21, 2021 11:17 pm
Chessqueen wrote: ↑Tue Sep 21, 2021 11:05 pm
lkaufman wrote: ↑Tue Sep 21, 2021 10:03 pm
Rebel wrote: ↑Tue Sep 21, 2021 9:16 pm
lkaufman wrote: ↑Tue Sep 21, 2021 8:35 pm
Rebel wrote: ↑Tue Sep 21, 2021 7:47 pm Knight odds matches

Komodo-Dragon-2
Code: Select all
Knight odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +344  =90 -266   *0  389.0   700   55.6%
  2 k2 099           +44  =14  -42   *0   51.0   100   51.0%
  3 Benjamin 1.0     +44  =11  -45   *0   49.5   100   49.5%
  4 ProDeo 2.2       +40  =15  -45   *0   47.5   100   47.5%
  5 Velvet 1.2.0     +40  =13  -47   *0   46.5   100   46.5%
  6 Dumb 1.8         +38  =14  -48   *0   45.0   100   45.0%
  7 Zahak 5.0        +36  =13  -51   *0   42.5   100   42.5%
  8 Fruit 2.1        +24  =10  -66   *0   29.0   100   29.0%

Total Games:     700
White Wins:      344 (49.1%)
Black Wins:      266 (38.0%)
Draws:            90 (12.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2757.5    51.0     100    51
   2 Komodo-Dragon 2    :  2750.5   389.0     700    56
   3 Benjamin 1.0       :  2747.0    49.5     100    50
   4 ProDeo 2.2         :  2732.9    47.5     100    48
   5 Velvet 1.2.0       :  2725.9    46.5     100    47
   6 Dumb 1.8           :  2715.3    45.0     100    45
   7 Zahak 5.0          :  2697.5    42.5     100    43
   8 Fruit 2.1          :  2593.5    29.0     100    29
Stockfish 14
Code: Select all
Knight odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14 +175  =50 -475   *0  200.0   700   28.6%
  2 ProDeo 2.2    +73  =11  -16   *0   78.5   100   78.5%
  3 Benjamin 1.0  +71  =10  -19   *0   76.0   100   76.0%
  4 Dumb 1.8      +73   =5  -22   *0   75.5   100   75.5%
  5 k2 099        +72   =6  -22   *0   75.0   100   75.0%
  6 Velvet 1.2.0  +67   =8  -25   *0   71.0   100   71.0%
  7 Fruit 2.1     +61   =7  -32   *0   64.5   100   64.5%
  8 Zahak 5.0     +58   =3  -39   *0   59.5   100   59.5%

Total Games:     700
White Wins:      175 (25.0%)
Black Wins:      475 (67.9%)
Draws:            50 (7.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 ProDeo 2.2      :  2798.5    78.5     100    79
   2 Benjamin 1.0    :  2773.5    76.0     100    76
   3 Dumb 1.8        :  2768.8    75.5     100    76
   4 k2 099          :  2764.1    75.0     100    75
   5 Velvet 1.2.0    :  2728.5    71.0     100    71
   6 Fruit 2.1       :  2676.2    64.5     100    65
   7 Zahak 5.0       :  2639.0    59.5     100    60
   8 Stockfish 14    :  2571.5   200.0     700    29
Komodo : 55.6%
Stockfish : 28.6%

Next, bishop-odds, same URL - http://rebel13.nl/a/grl.htm
Very nice! I assume since you don't say otherwise that you are using default Contempt for Komodo Dragon 2, which is just a nominal 8. I certainly can't complain about the results, but we would do even better with Contempt set to 100 or so. If this is accurate, then the lack on Contempt in Stockfish is no excuse for the poor relative performance, since Komodo's value is too low to add more than a percentage point or so to the results.
I am not very familiar with contempt settings, stockfish indeed has no such setting but likely it has some default setting in the source code. How do we get them equal?
Stockfish had a UCI option for Contempt but did away with it in SF14, though I don't know their reasons or whether they have a nonzero value set in the code. The Komodo Dragon value of 8 is much lower than what SF formerly used, but obviously higher than zero. If someone can verify that it does use zero in the code now, then you could lower the Dragon value from 8 to 0 to be equivalent, but that's a very minor change, hardly worth the bother. It might lower our winning percentage by one point or less, I would estimate.
The contempt value of Stockfish 13 is set to 24 on the configuration, I do NOT know if it is set to 24 in the code, either if by setting it to 8 will help stockfish 13 to be as strong as Komodo Dragon2 giving KInight, and Rook Odds.
My guess is that Stockfish13 with the biggest possible contempt is going to get a clearly better result than Stockfish14.

Probably so, but then Komodo Dragon2 should also be tested with Contempt 24 to be fair.

kyt_gg · Post by **kyt_gg** » Wed Sep 22, 2021 1:11 am

Hey Larry, don't know where to ask you this general question, but is Dragon also better against weaker opponents when compared to other engines that use MCTS (namely Leela)? I know how much you love MCTS.

I'm also going through your recently published book and am enjoying it!

Rebel · Post by **Rebel** » Wed Sep 22, 2021 2:07 am

BISHOP odds matches

Stockfish 14

Code: Select all

BISHOP odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14  +88  =27 -585   *0  101.5   700   14.5%
  2 Benjamin 1.0  +88   =4   -8   *0   90.0   100   90.0%
  3 Dumb 1.8      +85   =6   -9   *0   88.0   100   88.0%
  4 Velvet 1.2.0  +85   =5  -10   *0   87.5   100   87.5%
  5 Fruit 2.1     +83   =4  -13   *0   85.0   100   85.0%
  6 ProDeo 2.2    +84   =2  -14   *0   85.0   100   85.0%
  7 Zahak 5.0     +84   =1  -15   *0   84.5   100   84.5%
  8 k2 099        +76   =5  -19   *0   78.5   100   78.5%

Total Games:     700
White Wins:       88 (12.6%)
Black Wins:      585 (83.6%)
Draws:            27 (3.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 Benjamin 1.0    :  2824.1    90.0     100    90
   2 Dumb 1.8        :  2788.2    88.0     100    88
   3 Velvet 1.2.0    :  2780.1    87.5     100    88
   4 ProDeo 2.2      :  2743.1    85.0     100    85
   5 Fruit 2.1       :  2743.1    85.0     100    85
   6 Zahak 5.0       :  2736.3    84.5     100    85
   7 k2 099          :  2666.0    78.5     100    79
   8 Stockfish 14    :  2439.1   101.5     700    15

Komodo-Dragon-2

Code: Select all

BISHOP odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +284  =92 -324   *0  330.0   700   47.1%
  2 k2 099           +53  =16  -31   *0   61.0   100   61.0%
  3 Benjamin 1.0     +51  =15  -34   *0   58.5   100   58.5%
  4 ProDeo 2.2       +54   =9  -37   *0   58.5   100   58.5%
  5 Dumb 1.8         +43  =20  -37   *0   53.0   100   53.0%
  6 Velvet 1.2.0     +48  =10  -42   *0   53.0   100   53.0%
  7 Zahak 5.0        +41  =10  -49   *0   46.0   100   46.0%
  8 Fruit 2.1        +34  =12  -54   *0   40.0   100   40.0%

Total Games:     700
White Wins:      284 (40.6%)
Black Wins:      324 (46.3%)
Draws:            92 (13.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2775.7    61.0     100    61
   2 ProDeo 2.2         :  2757.5    58.5     100    59
   3 Benjamin 1.0       :  2757.5    58.5     100    59
   4 Velvet 1.2.0       :  2718.3    53.0     100    53
   5 Dumb 1.8           :  2718.3    53.0     100    53
   6 Komodo-Dragon 2    :  2697.3   330.0     700    47
   7 Zahak 5.0          :  2669.2    46.0     100    46
   8 Fruit 2.1          :  2626.2    40.0     100    40

Komodo : 47.1%
Stockfish : 14.5%

Tomorrow ROOK odds, same URL - http://rebel13.nl/a/grl.htm

lkaufman · Post by **lkaufman** » Wed Sep 22, 2021 2:43 am

Rebel wrote: ↑Wed Sep 22, 2021 2:07 am BISHOP odds matches

Stockfish 14

Code: Select all

BISHOP odds match Stockfish 14 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:
 
No. Name          Win Draw Loss Unf.  Score Games       %
---------------------------------------------------------
  1 Stockfish 14  +88  =27 -585   *0  101.5   700   14.5%
  2 Benjamin 1.0  +88   =4   -8   *0   90.0   100   90.0%
  3 Dumb 1.8      +85   =6   -9   *0   88.0   100   88.0%
  4 Velvet 1.2.0  +85   =5  -10   *0   87.5   100   87.5%
  5 Fruit 2.1     +83   =4  -13   *0   85.0   100   85.0%
  6 ProDeo 2.2    +84   =2  -14   *0   85.0   100   85.0%
  7 Zahak 5.0     +84   =1  -15   *0   84.5   100   84.5%
  8 k2 099        +76   =5  -19   *0   78.5   100   78.5%

Total Games:     700
White Wins:       88 (12.6%)
Black Wins:      585 (83.6%)
Draws:            27 (3.9%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER          :  RATING  POINTS  PLAYED   (%)
   1 Benjamin 1.0    :  2824.1    90.0     100    90
   2 Dumb 1.8        :  2788.2    88.0     100    88
   3 Velvet 1.2.0    :  2780.1    87.5     100    88
   4 ProDeo 2.2      :  2743.1    85.0     100    85
   5 Fruit 2.1       :  2743.1    85.0     100    85
   6 Zahak 5.0       :  2736.3    84.5     100    85
   7 k2 099          :  2666.0    78.5     100    79
   8 Stockfish 14    :  2439.1   101.5     700    15

Komodo-Dragon-2

Code: Select all

BISHOP odds match Komodo-Dragon-2 vs a pool of 2700-2730 elo rated engines
Time Control : Time control : 40/40
Games        : 700

Results from file all.pgn:

No. Name             Win Draw Loss Unf.  Score Games       %
------------------------------------------------------------
  1 Komodo-Dragon 2 +284  =92 -324   *0  330.0   700   47.1%
  2 k2 099           +53  =16  -31   *0   61.0   100   61.0%
  3 Benjamin 1.0     +51  =15  -34   *0   58.5   100   58.5%
  4 ProDeo 2.2       +54   =9  -37   *0   58.5   100   58.5%
  5 Dumb 1.8         +43  =20  -37   *0   53.0   100   53.0%
  6 Velvet 1.2.0     +48  =10  -42   *0   53.0   100   53.0%
  7 Zahak 5.0        +41  =10  -49   *0   46.0   100   46.0%
  8 Fruit 2.1        +34  =12  -54   *0   40.0   100   40.0%

Total Games:     700
White Wins:      284 (40.6%)
Black Wins:      324 (46.3%)
Draws:            92 (13.1%)
Unfinished:        0 (0.0%)

Estimated ratings for this elo 2715 pool

   # PLAYER             :  RATING  POINTS  PLAYED   (%)
   1 k2 099             :  2775.7    61.0     100    61
   2 ProDeo 2.2         :  2757.5    58.5     100    59
   3 Benjamin 1.0       :  2757.5    58.5     100    59
   4 Velvet 1.2.0       :  2718.3    53.0     100    53
   5 Dumb 1.8           :  2718.3    53.0     100    53
   6 Komodo-Dragon 2    :  2697.3   330.0     700    47
   7 Zahak 5.0          :  2669.2    46.0     100    46
   8 Fruit 2.1          :  2626.2    40.0     100    40

Komodo : 47.1%
Stockfish : 14.5%

Tomorrow ROOK odds, same URL - http://rebel13.nl/a/grl.htm

So bishops are indeed worth more than knights (at least when bishop pair is broken for the side losing the bishop), no surprise there. But it is interesting that Stockfish lost much more than Komodo from this, SF score was nearly cut in half going from knight odds to bishop odds! Regarding rook odds, it is roughly a class (200 elo) larger handicap than knight odds, so a field in the 2500 to 2530 range for opponents might be more balanced, but anyway it will be interesting.

Rebel · Post by **Rebel** » Wed Sep 22, 2021 7:15 am

lkaufman wrote: ↑Wed Sep 22, 2021 2:43 am So bishops are indeed worth more than knights (at least when bishop pair is broken for the side losing the bishop), no surprise there. But it is interesting that Stockfish lost much more than Komodo from this, SF score was nearly cut in half going from knight odds to bishop odds! Regarding rook odds, it is roughly a class (200 elo) larger handicap than knight odds, so a field in the 2500 to 2530 range for opponents might be more balanced, but anyway it will be interesting.

At the moment I am doing queen odds, just to be complete.

The rook epd is not good, see:

Code: Select all

rnbqkb1r/ppp1pppp/3p4/3nP3/3P4/8/PPP2PPP/1NBQKBNR w KQkq - 0 4; v=-526
r1bqkb1r/pppnpppp/3p1n2/8/2PP4/2N5/PP2PPPP/2BQKBNR w KQkq - 2 4; v=-529
rnbqk1nr/ppp1ppbp/3p2p1/8/3PP3/5N2/PPP2PPP/1NBQKB1R w KQkq - 0 4; v=-536
r1bqkb1r/ppp1pppp/2n2n2/3p4/2PP4/4P3/PP3PPP/1NBQKBNR w KQkq - 1 4; v=-538

Castling flags are wrong and positions are ignored by cute.

Does somebody has a good rook odds epd of (at least) 100 positions?

Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss

Re: Komodo-Dragon-2 vs Stockfish 14 at knight odss