K9 / SF dev. stats and analyzes ...

Frank Quisinsky · Post by **Frank Quisinsky** » Tue May 12, 2015 2:50 am

Hi there,

after my "Fight for Place 1" a bit to created stats!

1. Have a look in the simulated tournament table from my TOP-21. You can see that on first position is Komodo 9. Vs. stronger opponents Komodo seems a very bit stronger as SF. But I am thinking both engines are very closed.

http://www.amateurschach.de/ftptrigger/fct0-2015.html

2. Main apprehension with the new Komodo 9 I had are the first playing phase. Komodo 8 is very strong here and with improvements I life each time in danger that Komodo lost here playing strength for more power in the important playing phase (transposition into endgame).

But all is OK.
Komodo 9 don't lost here ... that is great!

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number 020-059
 Elostat 1.3,   EloØ = 2.898, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Komodo 9 x64                   : 3072  42 39  150  75.7%  2875  44.7%    80   67    3
 02. Stockfish 26.04.15 BMI2 x64    : 3043  33 30  180  71.4%  2884  56.1%    78  101    1

we can compare with the preview Versions of Komodo and Stockfish:
Nice to see all the draw games from Stockfish ... more clear this one if we are looking in the complete results, means the results vs. all 32 participants. I wrote it in the thread to my event.

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number 020-059
 Elostat 1.3,   EloØ = 2.896, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Komodo 8 x64                   : 3065  48 46  127  75.2%  2872  41.7%    69   53    5
 02. Stockfish 6 BMI2 x64           : 3032  30 27  208  71.4%  2873  57.2%    89  119    0

OK, so the improvments from Komodo should be to see in the other playing phases.

Let us look in the late middlegame and transposition endgame:

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number  060-099
 Elostat 1.3,   EloØ = 2.898, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Komodo 9 x64                   : 3198  32 31  600  87.6%  2859  19.2%   468  115   17
 02. Stockfish 26.04.15 BMI2 x64    : 3169  27 27  642  85.5%  2861  24.9%   469  160   13

And the preview versions:

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number  060-099
 Elostat 1.3,   EloØ = 2.896, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Stockfish 6 BMI2 x64           : 3185  31 30  582  86.0%  2870  21.1%   439  123   20
 02. Komodo 8 x64                   : 3141  29 28  601  82.5%  2871  23.3%   426  140   35

WOW !!
+ 58 Elo for Komodo 9 if we compare with Komodo 8 (same participants).
And clearly lesser are the lost games. This is a big jump here!! To see here again a lot of draw games from the current Stockfish version.

Normaly the endgame should go on Komodo and the transposition into endgame should go on Stockfish. That are my impressions during the "Fight for Place 1". But after my stats the situation is an other one! Means I am thinking during the fight that Komodo is very clearly improved in endgame and Stockfish is better in transposition into endgame.

We are so blind without such stats ...
Maybe I am blind ...

Attention:
In a Rating list with a lot of derviate and clones Komodo should for that reason here a small advantage to Stockfish. Think so ...

And what is happen in the endgame?
The big advantage from Stockfish to Komodo in the past!

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number 100-299
 Elostat 1.3,   EloØ = 2.898, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Stockfish 26.04.15 BMI2 x64    : 3047  34 32  178  68.3%  2914  55.6%    72   99    7
 02. Komodo 9 x64                   : 3031  27 26  250  66.8%  2909  59.2%    93  148    9

Code: Select all

 FCT1 (Frank's Continuous Tourney 0), games with move number 100-299
 Elostat 1.3,   EloØ = 2.896, Category = 25
     Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
 01. Stockfish 6 BMI2 x64           : 3046  32 30  210  69.8%  2901  53.8%    90  113    7
 02. Komodo 8 x64                   : 2981  23 22  272  61.4%  2901  67.6%    75  184   13

Again +50 Elo for Komodo 9 and now both engines are very closed too. Nice to see that Stockfish is faster here ... clearly lesser games goes in endgame with Stockfish because made more draw during the playing phases before. Komodo 9 hold in bad position more often a draw in endgames and won more games here.

All in all ...
Both engines are very closed with different styles. Stockfish try to find the faster way, often with draw and Komodo will see what is happen from the beginning and is more the classic fighter!

I like the style of Komodo more as the style of Stockfish. In the beginning very aggressive and later in the game Komodo is fighting for each advantage more as Stockfish which search more the fast way and if Stockfish can't find it ... draw is a good result in the first playing phase.

The complete stats to my TOP-21 can be found on my webpage.
I saw again enough from Komodo and Stockfish for the moment.

You can build what you like:
Place 1 for Stockfish dev. if more weaker engines are the participants.
Place 1 for Komodo if more stronger engines are the participants.

But right is ...
Very very closed and we can say ...
We have two engines on place 1!!

Best
Frank