Komodo 10.2 released.

[rcmaddox]

Purchased and downloaded Ok. The licensing string says it's an illegal copy.

See here -->> https://1drv.ms/i/s!AgVhuW8u-GbEg2URwSZMqtt6tZyN

[MikeB]

tc below = 15 sec /game/ 0.25 sec inc - my experience is the Komodo scales very well with a longer time control and more cores - so I would not read too much into the results below...

Code: Select all

200 rounds and 1200 games completed...
time control: 1500+25
Date: 10/29/16 : 22:12:56
1200 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 102916-14y       3152   0.0   14   14   600  357.0  59.5  200   86  314  33.3  52.3  3083 
   2 Stockfish 281016 64 POPC   3147   4.7   14   14   600  352.5  58.8  194   89  317  32.3  52.8  3084 
   3 Komodo 10.2 64-bit         3053  94.1   14   14   600  247.0  41.2  107  213  280  17.8  46.7  3116 
   4 Komodo 10.1 64-bit         3049   4.1   14   14   600  243.5  40.6  106  219  275  17.7  45.8  3117 
---------------------------------------------------------------------------------------------------------
1200 game(s) loaded

source for the modified version of stockfish 102916-14y can be found here...

https://github.com/MichaelB7/Stockfish and you will find a mac exe there as well along with the current cfish for the mac and current developmental stockfish for the mac.

[lkaufman]

tc below = 15 sec /game/ 0.25 sec inc - my experience is the Komodo scales very well with a longer time control and more cores - so I would not read too much into the results below...

Code: Select all

200 rounds and 1200 games completed...
time control: 1500+25
Date: 10/29/16 : 22:12:56
1200 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 102916-14y       3152   0.0   14   14   600  357.0  59.5  200   86  314  33.3  52.3  3083 
   2 Stockfish 281016 64 POPC   3147   4.7   14   14   600  352.5  58.8  194   89  317  32.3  52.8  3084 
   3 Komodo 10.2 64-bit         3053  94.1   14   14   600  247.0  41.2  107  213  280  17.8  46.7  3116 
   4 Komodo 10.1 64-bit         3049   4.1   14   14   600  243.5  40.6  106  219  275  17.7  45.8  3117 
---------------------------------------------------------------------------------------------------------
1200 game(s) loaded

source for the modified version of stockfish 102916-14y can be found here...

https://github.com/MichaelB7/Stockfish and you will find a mac exe there as well along with the current cfish for the mac and current developmental stockfish for the mac.

One reason we have made progress is that we have shifted away from relying too much on 15" +.1" games, which had been our main test level. Now we test much more at longer TC, and some changes were probably bad for 15" games but good for games with at least a couple minutes base time plus increment. So the small elo gain at this level doesn't worry me.

[mjlef]

Purchased and downloaded Ok. The licensing string says it's an illegal copy.

See here -->> https://1drv.ms/i/s!AgVhuW8u-GbEg2URwSZMqtt6tZyN

Robert,

We do not know what happened, but I just manually sent you another copy. Email us back if you have the same issue.

Mark

[lkaufman]

We have released komodo 10.2 at komodochess.com. We estimate based on our testing that it is about 22 elo stronger than komodo 10.1 and about 30 elo stronger than komodo 10 at three minutes plus one second increment, on one or many threads. This is the largest elo gain we have had between versions since going to the subscription model. There were so many changes that it is very difficult to say which ones should get most of the credit. Evaluation, search, and time management were all changed significantly. Features (UCI options) are the same as in komodo 10.1. As usual, it is free to subscribers, and available at a 20% discount to those who bought komodo 9 or a later version.

What _I_ want to see is how this new komodo 10.2 does vs the SF that played in stage 3 of TCEC, OR, the version of SF that played in Graham Banks matches between sf and komodo.

It is very frustrating to have two moving targets at the same time.

e.g. let's say komodo beats ASM, how will we know if the older SF version, which Graham used, wouldn't have crushed komodo 10.2 like it did to komodo 10.1?

We should first test a moving target with a standing target before we go on to the second moving target.

I think we'll do ok relative to stockfish in tests like Frank's where each opponent plays against many other opponents, but we will not look so good in direct matches with Stockfish. The reasons for this are not all clear, although "Contempt" is a factor. It's up to each person to decide whether a direct match or a rr tournament is a better way to decide which of two engines is stronger. There are arguments both ways.

[rcmaddox]

Thanks. Actually, the "illegal" copy works fine at full strength - but it's nice to have a personalized .exe.

I shouldn't have posted about this in the forum. My apologies to all.

[MikeB]

started a set with a longer time control :

Hardware: Mid 2010 Mac Pro, 12 Core, 64 GB RAM, os: Sierra 10.12.1
Bench: http://browser.primatelabs.com/v4/cpu/858736
time control: 3 min base with 3 second increment
threads: 2
hash: 512
positons: 50 - selected at random from a 2500 position base by a c-shell script, positions were culled from 40/40 CCRL games between opponents which were ranked over 3200 which ended in a non-draw and the score after move 8 was from 45 to 65 centipawns
games per position: 2 - alternating colors
engines: note: the "2C" below references that the ini file was setup with "Threads = 2"
set eng1 = komodo-10.1.2C.ini
set eng2 = Stockfish-102916-14y.2C.ini
set eng3 = Stockfish-281016.2C.ini
set eng4 = komodo-10.2.2C.ini
live updates: https://www.dropbox.com/s/q30k3kvul6qsa ... a.txt?dl=1 - every 12 games completed or about every 30 minutes or so
pgn file: https://www.dropbox.com/s/njswzxuo2u88gp2/all.pgn?dl=1
source: https://github.com/MichaelB7/Stockfish for the Stockfish fork "Stockfish-102916-14y" above

My experience is that this time control setting will result in a lot more draws than the micro time control setting.
first round is in:

Code: Select all

12 of 600 games completed...
time control: 18000+300
Date: 10/30/16 : 00:52:40
12 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Komodo 10.2 64-bit         3100   0.0    3    3     6    3.0  50.0    0    0    6   0.0 100.0  3100 
   2 Komodo 10.1 64-bit         3100   0.0    2    2     7    3.5  50.0    0    0    7   0.0 100.0  3100 
   3 Stockfish 102916-14y       3100   0.0    3    3     5    2.5  50.0    0    0    5   0.0 100.0  3100 
   4 Stockfish 281016 64 POPC   3100   0.0    3    3     6    3.0  50.0    0    0    6   0.0 100.0  3100 
---------------------------------------------------------------------------------------------------------
12 game(s) loaded

[MikeB]

after starting off with 18 straight draws, we do have our first decisive game:

[pgn][Event "Computer chess game"]
[Site "Mac-Pro"]
[Date "Sun Oct 30 01:10:06 2016"]
[Round "1"]
[White "Komodo 10.2 64-bit"]
[Black "Stockfish 281016 64 POPC"]
[Result "1-0"]
[TimeControl "18000+300"]
[FEN "r1bqr1k1/1pp2ppp/p1n2n2/3pp3/2PP4/P3PNP1/1P1NBPP1/R2QK2R w KQ - 0"]

1. c4d5 f6d5 2. d1c2 h7h6 3. d4e5 c6e5 4. f3e5 e8e5 5. e3e4 d5b6
6. e1g1 e5e6 7. a1d1 e6c6 8. c2b1 d8e7 9. d2b3 c8d7 10. e2f3
b6a4 11. b3a5 c6c5 12. b2b4 c5e5 13. a5c4 e5e6 14. d1c1 d7b5
15. f1d1 a8e8 16. c4e3 c7c6 17. e3f5 e7f8 18. f5d4 e6e5 19. d1e1
c6c5 20. d4b5 a6b5 21. b4c5 e5c5 22. e1e3 e8c8 23. c1e1 a4b6
24. e4e5 b6c4 25. e3e2 c4a3 26. b1a2 a3c2 27. e1d1 c2b4 28. a2b3
c5c3 29. b3b1 b4c6 30. f3d5 c6d4 31. e2d2 d4e6 32. b1b5 c8b8
33. b5b6 f8c5 34. b6b1 e6f8 35. d2d4 c5e7 36. d4f4 f8e6 37. f4b4
c3c7 38. b4b6 e6c5 39. b1a1 e7e8 40. b6d6 b7b5 41. a1a5 c7c8
42. a5a7 b5b4 43. d5c6 e8e5 44. d6d5 e5e2 45. a7c5 b4b3 46. d5d2
e2e6 47. d2d6 e6g4 48. c5d5 b3b2 49. d1b1 g4b4 50. d6d7 b4b3
51. g1h2 g8h8 52. d5b3 b8b3 53. c6d5 b3b6 54. d5f7 g7g6 55. g3g4
c8a8 56. f7d5 a8a5 57. f2f3 h6h5 58. h2g3 a5a1 59. d5e4 h5g4
60. g3g4 h8g8 61. g4f4 b6b4 62. d7d1 a1a5 63. d1d2 a5b5 64. d2d7
g6g5 65. f4e3 b5b6 66. e3d2 b6b5 67. g2g4 b5e5 68. b1h1 b2b1N
69. e4b1 b4b2 1-0 {Black resigns}[/pgn]

Look like the clinker move was 42. ... b5 which ultimately gave up the N, perhaps 42. ...Ne6 was better...
[d]1rr1q1k1/Q4pp1/3R3p/1pnBP3/8/6P1/5PP1/3R2K1 b - - 3 42

Code: Select all

24 of 600 games completed...
time control: 18000+300
Date: 10/30/16 : 01:17:23
24 game(s) loaded
Iteration 200: 4.82675e-05 
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 102916-14y       3116   0.0   40   40    12    6.5  54.2    1    0   11   8.3  91.7  3095 
   2 Komodo 10.1 64-bit         3101  14.4   43   43    12    6.0  50.0    1    1   10   8.3  83.3  3100 
   3 Stockfish 281016 64 POPC   3099   2.3   43   43    12    6.0  50.0    1    1   10   8.3  83.3  3100 
   4 Komodo 10.2 64-bit         3084  15.2   41   41    12    5.5  45.8    1    2    9   8.3  75.0  3105 
---------------------------------------------------------------------------------------------------------
24 game(s) loaded

[MikeB]

current results:

Code: Select all

36 of 600 games completed...
time control: 18000+300
Date: 10/30/16 : 01:42:35
36 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 102916-14y       3135   0.0   46   46    18   10.5  58.3    3    0   15  16.7  83.3  3088 
   2 Komodo 10.1 64-bit         3101  33.9   50   50    18    9.0  50.0    2    2   14  11.1  77.8  3100 
   3 Komodo 10.2 64-bit         3088  13.7   50   50    18    8.5  47.2    3    4   11  16.7  61.1  3104 
   4 Stockfish 281016 64 POPC   3076  11.3   49   49    18    8.0  44.4    1    3   14   5.6  77.8  3108 
---------------------------------------------------------------------------------------------------------
36 game(s) loaded

right after this round - K10.1 went on 3 game winning streak and is now leading...

[MikeB]

current results:

Code: Select all

60 of 600 games completed...
time control: 18000+300
Date: 10/30/16 : 02:28:38
60 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 102916-14y       3130   0.0   54   54    30   17.0  56.7    7    3   20  23.3  66.7  3090 
   2 Komodo 10.1 64-bit         3105  25.0   55   55    31   16.0  51.6    7    6   18  22.6  58.1  3099 
   3 Komodo 10.2 64-bit         3087  18.8   58   58    29   13.5  46.6    6    8   15  20.7  51.7  3104 
   4 Stockfish 281016 64 POPC   3077   9.2   58   58    30   13.5  45.0    6    9   15  20.0  50.0  3108 
---------------------------------------------------------------------------------------------------------
60 game(s) loaded

interestingly, under the current engagements rules of fishtesting - my changes will never pass fishtest - they only test at 10 sec with .1 sec which called STC or Short time Control or 60 seconds with .6 sec increment which is called LTC or Long Time Control. But one cannot argue with their success, but I do think they might be missing some opportunities which could only be gleaned from longer time controls.