Testing Komodo9: 1440 Games

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 9:09 pm

Re: Testing Komodo 9: 1440 games.

Post by Tomcass »

TESTING KOMODO 9.3: 1920 GAMES
FIRST LEG: 960 GAMES AT FIXED TIME CONTROL WITH TIME USAGE = 5


6 real cores
Ponder: Off.
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 512
Relative Speed: 29.54
Knodes per second: 14.177

Time control= 4 +0
Komodo 9.3 64-bit_FTC_TU5_X6 - Komodo 9.2 64-bit_x6 59.0 - 61.0 +12/=94/-14 49.17%
Komodo 9.3 64-bit_FTC_TU5_X6 - Stockfish 190915 64 POP 57.0 - 63.0 +20/=74/-26 47.50%
Komodo 9.3 64-bit_FTC_TU5_X6 - Houdini 4 x64_st_X6_CT0 86.0 - 34.0 +57/=58/-5 71.67%
Komodo 9.3 64-bit_FTC_TU5_X6 - Gull 3 x64 XP 80.5 - 39.5 +47/=67/-6 67.08%

Score after 480 games using 6 cores= 282.5 – 197.5= 58.86%
480 Games = http://www.mediafire.com/download/0y3c878h5m88r0r

8 real cores
Ponder: Off
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 1024
Relative Speed: 53.22
Knodes per Second: 25.543

Time Control= 4 + 0
Komodo 9.3 64-bit_FTC_TU5_X8 - Komodo 9.2 64-bit_x8 66.0 - 54.0 +21/=90/-9 55.00%
Komodo 9.3 64-bit_FTC_TU5_X8 - Stockfish 190915 64 BMI2_8 59.5 - 60.5 +20/=79/-21 49.58%
Komodo 9.3 64-bit_FTC_TU5_X8 - Houdini 4 Pro x64_Ct0_8 81.0 - 39.0 +52/=58/-10 67.50%
Komodo 9.3 64-bit_FTC_TU5_X8 - Gull 3 x64 XPx8 90.0 - 30.0 +64/=52/-4 75.00%

Score after 480 games using 8 cores= 295.5 – 183.5= 61.56%
480 Games = http://www.mediafire.com/download/t996hjk3ycq8j6p

Global score for Komodo 9.3 at Fixed Time Control with TIME USAGE = 5 after 960 games=
578.0 – 382.0= 60.21%

Average Elo of Oponents= 3.188
Estimated Elo for Komodo 9.3= 3.259

The score of Komodo 9.3 at fixed time control with time control= 5 has been 7 elo points better than Komodo 9.3 with default settings. If we compare this score with Komodo 9.0 at fixed time control to get a broader perspective we get:
Komodo 9.0 with Standard Settings = 3.204
Komodo 9.3 with Standard settings= 3.252 (16 Elo points average improvement per new release)
Komodo 9.3 with Time Usage 5 = 3.259. (18 Elo points average improvement per new release)

My tests confirm that the claim made by Komodo team that the average improvement of every new release is around 15 Elo points is true at Fixed Time Control.

I like this system that allows to enjoy a new Komodo every three months. My feeling is that I get value for my money.

I have started the second leg of this test of Komodo 9.3 (now standard settings) with Incremental Time Control.

Regards,

Tom.
Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 9:09 pm

Re: Testing Komodo 9: 1440 games.

Post by Tomcass »

TESTING KOMODO 9.3: 1920 GAMES
SECOND LEG: 960 GAMES AT INCREMENTAL TIME CONTROL WITH TIME USAGE = STANDARD


6 real cores
Ponder: Off.
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 512
Relative Speed: 29.54
Knodes per second: 14.177

Time Control= 2+2
Komodo 9.3 64-bit_nob_St_X6 - Komodo 9.2 64-bit_X6_nob 60.0 - 60.0 +12/=96/-12 50.00%
Komodo 9.3 64-bit_nob_St_X6 - Stockfish 190915 64 POPCNT_6 55.0 - 65.0 +18/=74/-28 45.83%
Komodo 9.3 64-bit_nob_St_X6 - Houdini 4 x64_st_X6_CT0 80.0 - 40.0 +50/=60/-10 66.67%
Komodo 9.3 64-bit_nob_St_X6 - Gull 3 x64 XP 84.0 - 36.0 +54/=60/-6 70.00%

Score after 480 games using 6 cores= 279.0 – 201.0= 58.12%
480 Games = http://www.mediafire.com/download/kcqzctaez50g4yl/

8 real cores
Ponder: Off
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 1024
Relative Speed: 53.22
Knodes per Second: 25.543

Time Control 2+2
Komodo 9.3 64-bit_St_nob_X8 - Komodo 9.2 64-bit_X6_nob 63.0 - 57.0 +15/=96/-9 52.50%
Komodo 9.3 64-bit_St_nob_X8 - Stockfish 190915 64 BMI2_8 64.0 - 56.0 +23/=82/-15 53.33%
Komodo 9.3 64-bit_St_nob_X8 - Houdini 4 Pro x64_Ct0_8 85.5 - 34.5 +54/=63/-3 71.25%
Komodo 9.3 64-bit_St_nob_X8 - Gull 3 x64 XPx8 84.5 - 35.5 +55/=59/-6 70.42%

Score after 480 games using 8 cores= 297.0 – 183.0 = 61.87%
480 Games= http://www.mediafire.com/download/ahftbwgqgxanxhw/

Global score for Komodo 9.3 at Incremental Time Control with TIME USAGE = 0 (Default) after 960 games= 576.0 – 384.0= 60.00%

Average Elo of Oponents= 3.188
Estimated Elo for Komodo 9.3= 3.258


--------------------------

GLOBAL SCORE FOR KOMODO 9.3 AFTER 1.920 GAMES= 1.154 – 766= 60.11%
Estimated Elo For Komodo 9.3 after 1.920 Games= 3.259


Error Bars +/-11

Ranking of selected engines under my testing conditions:

1.- Stockfish 121115 IPMAN= 3.277
2.- Stockfish 021115 KP= 3.267
3.- Stockfish Development 190915= 3.262
4.- Komodo 9.3 = 3.259
5.- Komodo 9.2 = 3.251.
6.- Komodo 9.1= 3.247
7.- Sugar Pro V.1= 3.246
8.- Stockfish 6= 3.219
9.- Komodo 9= 3.209
10.- Houdini 4= 3.136
11.- Gull 3 XP= 3.103

Regards from Barcelona.

Tom.
Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 9:09 pm

Re: Testing Komodo 9: 1440 games.

Post by Tomcass »

TESTING KOMODO 9.4: 960 GAMES

6 real cores
Ponder: Off.
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 512
Relative Speed: 29.54
Knodes per second: 14.177

Time control= 4 +0
Komodo 9.4 64-bit_X6_nob - SugaR 2.0 64 POPCNT_X6 28.0 - 32.0 +12/=32/-16 46.67%
Komodo 9.4 64-bit_X6_nob - Stockfish 020316 64 POPCNT_6 26.0 - 34.0 +8/=36/-16 43.33%
Komodo 9.4 64-bit_X6_nob - Houdini 4 Pro x64_St_Ct0_X6 37.5 - 22.5 +21/=33/-6 62.50%
Komodo 9.4 64-bit_X6_nob - Gull 3 x64 XP_X6 42.5 - 17.5 +31/=23/-6 70.83%

Time Control= 2+2
Komodo 9.4 64-bit_X6_nob - SugaR 2.0 64 POPCNT_X6 26.0 - 34.0 +9/=34/-17 43.33%
Komodo 9.4 64-bit_X6_nob - Stockfish 020316 64 POPCNT_6 26.0 - 34.0 +5/=42/-13 43.33%
Komodo 9.4 64-bit_X6_nob - Houdini 4 Pro x64_St_Ct0_X6 40.5 - 19.5 +24/=33/-3 67.50%
Komodo 9.4 64-bit_X6_nob - Gull 3 x64 XP_X6 40.5 - 19.5 +25/=31/-4 67.50%

Score after 480 games using 6 cores= 267.0 – 213.0= 55.62%
480 Games = http://www.mediafire.com/download/m4evp6qd98u5phi/

8 real cores
Ponder: Off
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 1024
Relative Speed: 53.22
Knodes per Second: 25.543

Time Control= 4 + 0
Komodo 9.4 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 25.5 - 34.5 +4/=43/-13 42.50%
Komodo 9.4 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 27.0 - 33.0 +10/=34/-16 45.00%
Komodo 9.4 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 38.0 - 22.0 +19/=38/-3 63.33%
Komodo 9.4 64-bit_X8_nob - Gull 3 x64 BMI2_X8 41.0 - 19.0 +26/=30/-4 68.33%

Time Control= 2+2
Komodo 9.4 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 27.0 - 33.0 +5/=44/-11 45.00%
Komodo 9.4 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 29.5 - 30.5 +9/=41/-10 49.17%
Komodo 9.4 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 38.5 - 21.5 +24/=29/-7 64.17%
Komodo 9.4 64-bit_X8_nob - Gull 3 x64 BMI2_X8 43.5 - 16.5 +29/=29/-2 72.50%

Score after 480 games using 8 cores= 270.0 – 210.0= 56.25%
Games= http://www.mediafire.com/download/q26lrcoq5oqbk3n

Global Score for Komodo 9.4 after 960 Games= 537.0 – 423.0= 55.94%
Average Elo of Oponents= 3203
Estimated Elo for Komodo 9.4 after 960 Games= 3245


I know that this test is useless because of the bugs found in Komodo 9.4. The score has been 14 elo points below Komodo 9.3. Anyway I have decided to post it including the games.

This afternoon I have started the 1920 Games test of Komodo 9.42 … and the score after 160 games is good so far, with a projection around 15 Elo points better than Komodo 9.3, as expected by Komodo’s programmers.

Let’s see the final score after 1920 games.

Regards,

Tom.
Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 9:09 pm

Re: Testing Komodo 9: 1440 games.

Post by Tomcass »

TESTING KOMODO 9.42 1.920 GAMES

6 real cores
Ponder: Off.
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 512
Relative Speed: 29.54
Knodes per second: 14.177

Time control= 4 +0
Komodo 9.42 64-bit_nob_X6 - SugaR 2.0 64 POPCNT_X6 33.0 - 27.0 +15/=36/-9 55.00%
Komodo 9.42 64-bit_nob_X6 - Stockfish 020316 64 POPCNT_6 27.5 - 32.5 +9/=37/-14 45.83%
Komodo 9.42 64-bit_nob_X6 - Houdini 4 Pro x64_St_Ct0_X6 41.0 - 19.0 +29/=24/-7 68.33%
Komodo 9.42 64-bit_nob_X6 - Gull 3 x64 XP_X6 42.5 - 17.5 +30/=25/-5 70.83%

Komodo 9.42 64-bit_nob_X6 - SugaR 2.0 64 POPCNT_X6 28.0 - 32.0 +10/=36/-14 46.67%
Komodo 9.42 64-bit_nob_X6 - Stockfish 020316 64 POPCNT_6 27.0 - 33.0 +8/=38/-14 45.00%
Komodo 9.42 64-bit_nob_X6 - Houdini 4 Pro x64_St_Ct0_X6 41.5 - 18.5 +27/=29/-4 69.17%
Komodo 9.42 64-bit_nob_X6 - Gull 3 x64 XP_X6 43.0 - 17.0 +30/=26/-4 71.67%

Time Control= 2+2
Komodo 9.42 64-bit_nob_X6 - SugaR 2.0 64 POPCNT_X6 27.5 - 32.5 +10/=35/-15 45.83%
Komodo 9.42 64-bit_nob_X6 - Stockfish 020316 64 POPCNT_6 27.5 - 32.5 +8/=39/-13 45.83%
Komodo 9.42 64-bit_nob_X6 - Houdini 4 Pro x64_St_Ct0_X6 46.0 - 14.0 +35/=22/-3 76.67%
Komodo 9.42 64-bit_nob_X6 - Gull 3 x64 XP_X6 40.0 - 20.0 +23/=34/-3 66.67%

Komodo 9.42 64-bit_nob_X6 - SugaR 2.0 64 POPCNT_X6 23.0 - 37.0 +4/=38/-18 38.33%
Komodo 9.42 64-bit_nob_X6 - Stockfish 020316 64 POPCNT_6 30.5 - 29.5 +9/=43/-8 50.83%
Komodo 9.42 64-bit_nob_X6 - Houdini 4 Pro x64_St_Ct0_X6 38.0 - 22.0 +21/=34/-5 63.33%
Komodo 9.42 64-bit_nob_X6 - Gull 3 x64 XP_X6 39.0 - 21.0 +23/=32/-5 65.00%

Score after 960 games using 6 cores= 555.0 – 405.0 = 57.81%
960 Games = http://www.mediafire.com/download/sw33ivbusjhqya1

8 real cores
Ponder: Off
GUI: Fritz 14
Book: Perfect 2014c – Limit 5 moves. Sedat –
No tablebases. No RTB used.
Hash 1024
Relative Speed: 53.22
Knodes per Second: 25.543

Time Control= 4 + 0
Komodo 9.42 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 29.5 - 30.5 +8/=43/-9 49.17%
Komodo 9.42 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 26.0 - 34.0 +7/=38/-15 43.33%
Komodo 9.42 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 44.5 - 15.5 +31/=27/-2 74.17%
Komodo 9.42 64-bit_X8_nob - Gull 3 x64 BMI2_X8 43.0 - 17.0 +30/=26/-4 71.67%

Komodo 9.42 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 30.5 - 29.5 +13/=35/-12 50.83%
Komodo 9.42 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 24.5 - 35.5 +5/=39/-16 40.83%
Komodo 9.42 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 43.5 - 16.5 +28/=31/-1 72.50%
Komodo 9.42 64-bit_X8_nob - Gull 3 x64 BMI2_X8 44.0 - 16.0 +31/=26/-3 73.33%

Time Control= 2+2
Komodo 9.42 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 26.5 - 33.5 +5/=43/-12 44.17%
Komodo 9.42 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 28.5 - 31.5 +9/=39/-12 47.50%
Komodo 9.42 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 39.0 - 21.0 +21/=36/-3 65.00%
Komodo 9.42 64-bit_X8_nob - Gull 3 x64 BMI2_X8 43.0 - 17.0 +28/=30/-2 71.67%

Komodo 9.42 64-bit_X8_nob - SugaR 2.0 64 BMI2_X8 28.0 - 32.0 +8/=40/-12 46.67%
Komodo 9.42 64-bit_X8_nob - Stockfish 020316 64 BMI2_8 25.0 - 35.0 +5/=40/-15 41.67%
Komodo 9.42 64-bit_X8_nob - Houdini 4 Pro x64_Ct0_8 46.5 - 13.5 +33/=27/-0 77.50%
Komodo 9.42 64-bit_X8_nob - Gull 3 x64 BMI2_X8 43.0 - 17.0 +29/=28/-3 71.67%

Score after 960 games using 8 cores= 565.0 – 395.0 = 58.85%
960 Games= http://www.mediafire.com/download/y1jqo9hck5r9cyw

Global Score for Komodo 9.42 after 1920 Games= 1.120 – 800.0 = 58.33%
Against
SugaR 2.0 (3285) = 47.08%
Stockfish 020316 (3287) = 45.10%
Houdini 4 Pro (3136) = 70.83%
Gull 3 XP (3103) = 70.31%


Average Elo of Oponents= 3203
Estimated ELO for Komodo 9.42 after 1.920 Games = 3261


In this test, the improvement of Komodo 9.42 over Komodo 9.3 is only 2 Elo points, obviously within error bars.


RANKING OF SELECTED ENGINES

1.- Stockfish Development 020316= 3.287
2.- SugaR 2.0= 3.285
3.- Stockfish Development 150116 IPMAN= 3.283
4.- SugaR Pro v1.3= 3.280
5.- Stockfish 7 = 3.268
6.- Komodo 9.42= 3.261
7.- Komodo 9.3 = 3.259
8.- Komodo 9.2 = 3.251.
9.- Komodo 9.1= 3.247
10.- DON 070316 = 3.246
11.- Stockfish 6= 3.219
12.- Komodo 9= 3.209
13.- Houdini 4= 3.136
14.- Gull 3 Development 150216= 3.120
15.- Gull 3 XP= 3.103

Regards,

Tom.