"The Fight for Place 1" ... will be start soon!

Frank Quisinsky · Post by **Frank Quisinsky** » Fri May 08, 2015 8:23 pm

Hi there,

with the same quantity of draw games compare to Komodo 9 and 5 more lost games (ratio to draw games with the other games I have) Stockfish can be after my short calculation 16 Elo stronger!

In this case can be the different between Komodo 9 and Stockfish dev. exactly 20 Elo (but only in this case). Menas SF dev. can be 20 Elo stronger as K9.

Enough ...
All what I saw is the same as I await.
In this case all is very boring ... but my book is clearly improved.

Thanks for it ... good for all others participants in my list.

Which such a bad stat:
Comparing this two engines a GM would never, never, never used Stockfish for opening book analyzes. Each time Komodo 9 is here clearly more interesting for each strong player !!!

This stat is a disaster from an engine with such a high playing strength.

Best
Frank

beram · Post by **beram** » Fri May 08, 2015 9:27 pm

Frank Quisinsky wrote:Short Hint:
If I delete 21 draw games from Stockfish ... K9 and SF have the same quantiy of draw games (9 games) ... we can see better the Elo difference between this engines.

I think all the draw games is a big problem in SF development. With better optimations you can see very easy how strong SF can be ...
Code: Select all
   # PLAYER                           &#58; RATING  ERROR   POINTS  PLAYED    (%)
   1 Stockfish 26.04.15 BMI2 x64      &#58; 3170.9   34.3    345.0     408   84.6%
   2 Komodo 9 x64                     &#58; 3142.1   32.3    353.0     429   82.3%
   3 Stockfish 6 BMI2 x64             &#58; 3135.5   16.6   1424.0    1700   83.8%
   4 Komodo 8 x64                     &#58; 3105.3   15.6   1375.5    1700   80.9%
We can see the different is 28,8 Elo and not 9 Elo.

With others words ...
Komodo can make +20 Elo with avoid draw games.

OK, the question is ...
If I avoid draw more lost games will be the result. So I must calculate / simulate all in ratio to more lost games. In this case I must added to the simulation after I delete 21 draw games 2 lost games (will be the ratio result). The differenz is 23 Elo and the final result is the same as with SF predecessor versions ... again and again SF lost 15 Elo.

+ 15 Elo if SF comes with better standard settings in avoid fast draw games.
That result is since a while allways the same. And I am not happy with the contemp settings SF have. I try it out for a while with Contemp = 17 and I can't see a better result as without Contemp.

Best
Frank

Without all the draw games it seems SF actual dev. is around 25 Elo stronger. This should around the results in testing with stronger opponents only ... maybe 20 Elo without to chance anything with avoid short draw games.

Hi Frank,

Why not just simply look at the games they played against each other ?
A match result of 57% is telling enough, SF dev is substantially stronger than Komodo 9:

Code: Select all

Stockfish 26.04.15 BMI2 x64 - Komodo 9 x64 &#40;3150&#41;		24.5	-	18.5		56.98%

Also in Graham Banks CCRL Komodo 9 64-bit 1CPU gauntlet, Sf6 is leading in its 'privat' match with 7(!) games to 1 (5 draws)

Komodo 9 is great improvement (against various engines) but in its matches against Stockfish the latter is still stronger

Frank Quisinsky · Post by **Frank Quisinsky** » Fri May 08, 2015 10:55 pm

Hi Bram,

I am looking deeper ...
Not on one result only, that's not interesting!!

But if you are looking on one result ...

Two round robins and two match results between Komodo and Stockfish. 100 games are to Play, 2x 50 ...

Current results here:

SF dev. round robin:
Stockfish 26.04.15 BMI2 x64 - Komodo 9 x64 (3150) 25.0 - 19.0 56.82% Perf=3197

Komodo 9 round-robin:
Komodo 9 x64 - Stockfish 26.04.15 BMI2 x64 (3150) 23.5 - 19.5 54.65% Perf=3182

In this case at the moment:
SF dev. vs. K9 = 44.5 : 42.5

Best
Frank

Frank Quisinsky · Post by **Frank Quisinsky** » Sat May 09, 2015 6:59 pm

Code: Select all

 FCP version     &#58; 2.07.1485.s7+s8
 Time            &#58; April 28th 2015 - May 11th, 2015
 Games           &#58; after 2880 of 3.200 &#40;4.357Mb&#41;
 Performance     &#58; Stockfish 26.04.15 BMI2 x64 = +13 Elo
 Performance     &#58; Komodo 9 x64                = +41 Elo


 Calculated with Ordo 1.0
 Stats after 100, 200 ... games!       Elo   Games  Score  Draws   White Black Points   w/ d/ l
 01. Stockfish 6 BMI2 x64              3135  1.700  Predecessor
 --. Stockfish 260415 BMI2 x64         ----  1.650  --.-%  --.-%   --,-  --,-  ---,-   ---/--/--
 01. Stockfish 260415 BMI2 x64         3148  1.485  83.0%  30.5%   70,0  67,5  137,5   114/47/04
 01. Stockfish 260415 BMI2 x64         3147  1.320  83.0%  30.8%   70,5  57,0  137,5   112/51/02
 01. Stockfish 260415 BMI2 x64         3146  1.155  82.9%  30.7%   70,0  64,5  134,5   106/57/02
 01. Stockfish 260415 BMI2 x64         3150    990  83.1%  30.1%   70,0  66,5  136,5   110/53/02
 01. Stockfish 260415 BMI2 x64         3150    825  83.2%  29.7%   72,0  65,5  137,5   116/43/06
 01. Stockfish 260415 BMI2 x64         3150    660  83.2%  30.6%   72,5  64,0  136,5   110/53/02
 01. Stockfish 260415 BMI2 x64         3152    495  83.3%  30.1%   71,5  66,5  138,0   114/48/03
 01. Stockfish 260415 BMI2 x64         3150    330  83.2%  30.6%   73,5  63,5  137,0   111/52/02
 01. Stockfish 260415 BMI2 x64         3150    165  83.3%  29.7%   73,5  64,0  137,5   113/49/03

 Calculated with Ordo 1.0
 Stats after 100, 200 ... games!       Elo   Games  Score  Draws   White Black Points   w/ d/ l
 02. Komodo 8 x64                      3105  1.700  Predecessor
 --. Komodo 9 x64                      ----  1.650  --.-%  --.-%   --,-  --,-  ---,-   ---/--/--
 01. Komodo 9 x64                      3145  1.485  82.8%  29.2%   72,5  64,0  136,5   111/51/03
 02. Komodo 9 x64                      3145  1.320  82.8%  29.0%   73,0  67,5  140,5   121/39/05
 02. Komodo 9 x64                      3141  1.155  82.4%  29.8%   70,5  64,0  134,5   108/53/04
 02. Komodo 9 x64                      3143    990  82.6%  29.4%   72,0  68,0  140,0   118/44/03
 02. Komodo 9 x64                      3137    825  82.1%  29.9%   73,5  61,5  135,5   110/50/05
 02. Komodo 9 x64                      3138    660  82.2%  29.8%   72,5  65,5  138,0   112/52/01
 02. Komodo 9 x64                      3132    495  81.7%  29.3%   71,0  62,5  133,5   110/47/08
 02. Komodo 9 x64                      3137    330  82.1%  29.7%   73,5  65,0  138,5   115/47/03
 02. Komodo 9 x64                      3114    165  80.3%  30.9%   73,5  59,0  132,5   107/51/07
  

     Program                           Elo    +    -  Games   Score  Av.Op.  Draw
 01. Stockfish 26.04.15 BMI2 x64       3148   18   18  1485   83.0%   2842   30.5%
 02. Komodo 9 x64                      3145   17   17  1485   82.8%   2842   29.2%
 --. Stockfish 6 BMI2 x64              3135   16   16  1700   83.8%   2823   28.9%
 --. Komodo 8 x64                      3104   16   16  1700   80.9%   2830   31.6%
 03. Fire 4 x64                        3045   14   14  1940   74.6%   2833   34.6%
 04. GullChess 3.0 BMI2 x64            3039   14   14  1790   72.8%   2847   40.1%
 05. Equinox 3.30 x64                  2991   12   12  1940   69.0%   2835   41.5%
 06. Sting SF 4.8.4 x64                2933   12   12  1840   61.5%   2844   44.6%
 07. Protector 1.7.0 x64               2910   13   13  1640   56.3%   2865   44.9%
 08. Critter 0.90 SSE4 x64             2899   13   13  1740   56.1%   2854   45.2%
 09. Hannibal 1.5 x64                  2896   12   12  1790   56.9%   2846   45.0%
 09. Texel 1.05 x64                    2896   12   12  1840   57.0%   2844   43.2%
 11. Chiron 2.0 x64                    2895   10   10  2890   61.7%   2806   40.4%
 12. Naum 4.6 x64                      2884   12   12  2090   58.0%   2824   42.8%
 --. Hannibal 1.4b x64                 2862   10   10  2500   59.2%   2794   43.3%
 --. Texel 1.04 x64                    2847   11   11  2300   57.5%   2793   42.4%
 13. Nirvanachess 2.0a x64             2837   11   11  2140   52.4%   2823   45.2%
 14. Senpai 1.0 SSE42 x64              2829    9    9  2890   53.3%   2808   42.4%
 15. Hiarcs 14 WCSC w32                2825   10   10  2890   52.7%   2809   43.0%
 16. Andscacs 0.72 POP x64             2813   11   11  2090   49.1%   2826   42.7%
 17. Sjeng c't 2010 w32                2796   12   12  1990   46.4%   2830   42.2%
 18. Shredder 12 x64                   2794   12   12  1790   43.0%   2854   42.3%
 19. Junior 13.3.00 x64                2783   10   10  2740   46.4%   2816   40.2%
 20. Spike 1.4 Leiden w32              2773   12   12  1690   39.3%   2863   41.9%
 21. Quazar 0.4 x64                    2758   14   14  1590   38.1%   2859   39.9%

Frank Quisinsky · Post by **Frank Quisinsky** » Sat May 09, 2015 7:21 pm

Code: Select all

Stockfish 26.04.15 BMI2 x64  = 1252.0 - 234.0  84.25%  Perf=3125
Komodo 9 x64                 = 1248.5 - 237.5  84.02%  Perf=3122

114 games for K9 and SF dev. to play only and the different is 3.5 points only. All is possible ...

Best
Frank

Frank Quisinsky · Post by **Frank Quisinsky** » Mon May 11, 2015 10:55 am

Hi there,

here are the final results:
http://www.amateurschach.de/ftptrigger/ ... 2-x64.html
http://www.amateurschach.de/ftptrigger/ ... 9-x64.html

Stockfish 26.04.15 BMI2 x64 - Komodo 9 x64 = 51.0 : 49.0

- games are online
- new rating list is online
- stats follow in the night (at the moment no time for it).
- my "FCP Live Book" is clearly improved with the new 3200 games. New version is online.

But ...
In my TOP-21 list Komodo 9 is 7 points better after 1.000 games as Stockfish 26.04.15 BMI2 x64. Clear, Stockfish lost to many games vs. weaker opponents with fast draw games.

In my opinion the more interesting engine is: Komodo 9 x64. The playing strength to Stockfish 26.04.15 BMI2 x64 is around the same. But this one was clear before I start this fight, also that the result is 10 Elo lesser as from others testers (my conditions). All in all the results are the same because Komodo 9 is around 10 Elo better as Stockfish 6, same if I looking in the results by others.

Later ...

Code: Select all

 FCP version     &#58; 2.07.s8
 Time            &#58; April 28th 2015 - May 10th, 2015
 Performance     &#58; Stockfish 26.04.15 BMI2 x64 = +15 Elo
 Performance     &#58; Komodo 9 x64                = +42 Elo


 Calculated with Ordo 1.0
 Stats after 100, 200 ... games!       Elo   Games  Score  Draws   White Black Points   w/ d/ l
 01. Stockfish 6 BMI2 x64              3134  1.700  Predecessor
 01. Stockfish 26.04.15 BMI2 x64       3149  1.650  83.1%  30.5%   72,0  67,0  139,0   114/50/01
 01. Stockfish 26.04.15 BMI2 x64       3148  1.485  83.0%  30.5%   70,0  67,5  137,5   114/47/04
 01. Stockfish 26.04.15 BMI2 x64       3147  1.320  83.0%  30.8%   70,5  57,0  137,5   112/51/02
 01. Stockfish 26.04.15 BMI2 x64       3146  1.155  82.9%  30.7%   70,0  64,5  134,5   106/57/02
 01. Stockfish 26.04.15 BMI2 x64       3150    990  83.1%  30.1%   70,0  66,5  136,5   110/53/02
 01. Stockfish 26.04.15 BMI2 x64       3150    825  83.2%  29.7%   72,0  65,5  137,5   116/43/06
 01. Stockfish 26.04.15 BMI2 x64       3150    660  83.2%  30.6%   72,5  64,0  136,5   110/53/02
 01. Stockfish 26.04.15 BMI2 x64       3152    495  83.3%  30.1%   71,5  66,5  138,0   114/48/03
 01. Stockfish 26.04.15 BMI2 x64       3150    330  83.2%  30.6%   73,5  63,5  137,0   111/52/02
 01. Stockfish 26.04.15 BMI2 x64       3150    165  83.3%  29.7%   73,5  64,0  137,5   113/49/03

 Calculated with Ordo 1.0
 Stats after 100, 200 ... games!       Elo   Games  Score  Draws   White Black Points   w/ d/ l
 02. Komodo 8 x64                      3104  1.700  Predecessor
 02. Komodo 9 x64                      3146  1.650  82.8%  29.2%   71,0  67,0  138,0   114/48/03
 02. Komodo 9 x64                      3145  1.485  82.8%  29.2%   72,5  64,0  136,5   111/51/03
 02. Komodo 9 x64                      3145  1.320  82.8%  29.0%   73,0  67,5  140,5   121/39/05
 02. Komodo 9 x64                      3141  1.155  82.4%  29.8%   70,5  64,0  134,5   108/53/04
 02. Komodo 9 x64                      3143    990  82.6%  29.4%   72,0  68,0  140,0   118/44/03
 02. Komodo 9 x64                      3137    825  82.1%  29.9%   73,5  61,5  135,5   110/50/05
 02. Komodo 9 x64                      3138    660  82.2%  29.8%   72,5  65,5  138,0   112/52/01
 02. Komodo 9 x64                      3132    495  81.7%  29.3%   71,0  62,5  133,5   110/47/08
 02. Komodo 9 x64                      3137    330  82.1%  29.7%   73,5  65,0  138,5   115/47/03
 02. Komodo 9 x64                      3114    165  80.3%  30.9%   73,5  59,0  132,5   107/51/07
  

     Program                           Elo    +    -  Games   Score  Av.Op.  Draw
 01. Stockfish 26.04.15 BMI2 x64       3149   16   16  1650   83.1%   2842   30.5%
 02. Komodo 9 x64                      3146   16   16  1650   82.8%   2842   29.2%
 --. Stockfish 6 BMI2 x64              3134   17   17  1700   83.8%   2823   28.9%
 --. Komodo 8 x64                      3104   15   15  1700   80.9%   2829   31.6%
 03. Fire 4 x64                        3045   13   13  1950   74.4%   2835   34.8%
 04. GullChess 3.0 BMI2 x64            3038   14   14  1800   72.5%   2849   39.9%
 05. Equinox 3.30 x64                  2991   13   13  1950   68.8%   2836   41.6%
 06. Sting SF 4.8.4 x64                2933   12   12  1850   61.3%   2846   44.6%
 07. Protector 1.7.0 x64               2910   13   13  1650   56.1%   2866   44.9%
 08. Critter 0.90 SSE4 x64             2899   12   12  1750   55.9%   2856   45.1%
 09. Hannibal 1.5 x64                  2896   12   12  1800   56.7%   2847   44.9%
 10. Texel 1.05 x64                    2895   12   12  1850   56.7%   2846   43.0%
 10. Chiron 2.0 x64                    2895    9    9  2900   61.5%   2807   40.3%
 12. Naum 4.6 x64                      2884   11   11  2100   57.8%   2825   42.8%
 --. Hannibal 1.4b x64                 2862   11   11  2500   59.2%   2794   43.3%
 --. Texel 1.04 x64                    2847   11   11  2300   57.5%   2792   42.4%
 13. Nirvanachess 2.0a x64             2837   11   11  2150   52.2%   2825   45.1%
 14. Senpai 1.0 SSE42 x64              2829    9    9  2900   53.2%   2809   42.4%
 15. Hiarcs 14 WCSC w32                2825    9    9  2900   52.6%   2810   43.0%
 16. Andscacs 0.72 POP x64             2813   11   11  2100   48.9%   2827   42.6%
 17. Sjeng c't 2010 w32                2796   12   12  2000   46.2%   2832   42.1%
 18. Shredder 12 x64                   2794   12   12  1800   42.8%   2856   42.2%
 19. Junior 13.3.00 x64                2783   10   10  2750   46.3%   2817   40.2%
 20. Spike 1.4 Leiden w32              2772   13   13  1700   39.1%   2865   41.7%
 21. Quazar 0.4 x64                    2758   13   13  1600   38.0%   2861   39.8%

egiovannotti · Post by **egiovannotti** » Tue May 12, 2015 1:28 am

Where can be download the games?

Graham Banks · Post by **Graham Banks** » Tue May 12, 2015 1:38 am

beram wrote:... in Graham Banks CCRL Komodo 9 64-bit 1CPU gauntlet, Sf6 is leading in its 'privat' match with 7(!) games to 1 (5 draws)

It's actually 6-1 with 7 draws so far in the gauntlet I'm running.
However, to balance that out a little, Komodo 9 just defeated Stockfish 6 by 12.5-11.5 in the final of the matchplay tournament that I ran.
These two engines are very close in strength, with Komodo slightly ahead in the rating lists.

Frank Quisinsky · Post by **Frank Quisinsky** » Tue May 12, 2015 1:43 am

Hi Giovannotti,

two possibilities:

1. Download selection, complete documentation with games in *.pgn.
http://www.amateurschach.de/main/_download.htm
Current database version: 2.07.s8

If you have interest on the complete database and much others files I created (documentation) please donwload the complete database (around 125Mb).

2. Download games by players.
Here you can find for each participant engine the white / black games in *.pgn if you like (programmers selection).
http://www.amateurschach.de/main/_sgbp.htm

Best
Frank

egiovannotti · Post by **egiovannotti** » Tue May 12, 2015 2:18 am

Frank Quisinsky wrote:Hi Giovannotti,

two possibilities:

1. Download selection, complete documentation with games in *.pgn.
http://www.amateurschach.de/main/_download.htm
Current database version: 2.07.s8

If you have interest on the complete database and much others files I created (documentation) please donwload the complete database (around 125Mb).

2. Download games by players.
Here you can find for each participant engine the white / black games in *.pgn if you like (programmers selection).
http://www.amateurschach.de/main/_sgbp.htm

Best
Frank

Thanks Frank!

"The Fight for Place 1" ... will be start soon!

Re: 16Elo points goes in Nirvana, not Nirvanachess means!

Re: 15 Elo lesser as SF actually is ...

Re: Correction ...

Re: After 2880 of 3.200 games = SFdev +13 Elo, K9 +41 Elo

Re: 3.5 points different only, all is possible!

Re: Final results = SFdev +15 Elo, K9 +42 Elo

Re: "The Fight for Place 1" ... will be start soon

Re: 15 Elo lesser as SF actually is ...

Re: Games ...

Re: Games ...