Testing Stockfish 11-03-13. 480 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: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

TESTING STOCKFISH DEVELOPMENT 121015 = 1920 GAMES

Timestamp: 1444683654 Bench: 7677367

SECOND LEG. INCREMENTAL TIME CONTROL. 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

Stockfish 121015 64 POPCNT_X6 - Komodo 9.2 64-bit_x6 61.0 - 59.0 +21/=80/-19 50.83%
Stockfish 121015 64 POPCNT_X6 - Stockfish 190915 64 POP_6 58.0 - 62.0 +8/=100/-12 48.33%
Stockfish 121015 64 POPCNT_X6 - Houdini 4 x64_st_X6_CT0 80.5 - 39.5 +48/=65/-7 67.08%
Stockfish 121015 64 POPCNT_X6 - Gull 3 x64 XP 83.5 - 36.5 +52/=63/-5 69.58%

Score using 6 cores: = 283.0 – 197.0= 58.96%
480 Games: http://www.mediafire.com/download/yt6rwrif44r3w64

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

Stockfish 121015 64 BMI2_X8 - Komodo 9.2 64-bit_x8 62.0 - 58.0 +22/=80/-18 51.67%
Stockfish 121015 64 BMI2_X8 - Stockfish 190915 64 BMI2_8 61.0 - 59.0 +7/=108/-5 50.83%
Stockfish 121015 64 BMI2_X8 - Houdini 4 Pro x64_Ct0_8 72.0 - 48.0 +31/=82/-7 60.00%
Stockfish 121015 64 BMI2_X8 - Gull 3 x64 XPx8 77.0 - 43.0 +40/=74/-6 64.17%

Score using 8 real cores: = 272.0 – 208.0 = 56.67%
480 Games: http://www.mediafire.com/download/xlrly3r3d1qmuwu
Against : Komodo 9.2 (3251) = 51.25% Stockfish 190915 (3262) = 49.58% Houdini 4 (3136) = 63.54%, Gull 3 XP (3103) = 66.87%
Average Elo of Oponents= 3.188

GLOBAL SCORE AFTER 960 GAMES AT INCREMENTAL TIME CONTROL= 555.0 – 405.0= 57.81%

ESTIMATED ELO FOR STOCKFISH DEV 121015 AT INCREMENTAL TIME CONTROL AFTER 960 GAMES= 3.243

================================

GLOBAL SCORE FOR STOCKFISH DEV 121015 AFTER 1.920 GAMES = 1.131 – 789.0= 58.91%
ESTIMATED GLOBAL ELO FOR STOCKFISH DEV 121015 AFTER 1.920 GAMES= 3.250

Error Bars +/- 11

Stockfish Development 121015 has scored 12 Elo Point below the current leader in my tests STOCKFISH DEVELOPMENT 190915 (3.262).

Let’s compare the latest scores by Stockfish Development and compiles since the record by SF Dev 190915:

STOCKFISH DEVELOPMENT 190915 = 3.262.
STOCKFISH BULLET 280915= 3.259
STOCKFISH DEVELOPMENT 031015= 3.247
STOCKFISH 061015 VOKAVOR = 3.256
STOCKFISH DEVELOPMENT 121015= 3.250

Regards,

Tom.
jarkkop
Posts: 198
Joined: Thu Mar 09, 2006 2:44 am
Location: Helsinki, Finland

Re: For Tom- or Maybe Anyone Who Knows!!

Post by jarkkop »

stockfish_15102104_x64_modern has done a new record in my small test.
No losses to Houdini3. 30 matches and SF got 88% of points.
Give it a chance.

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

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

jarkkop wrote:stockfish_15102104_x64_modern has done a new record in my small test.
No losses to Houdini3. 30 matches and SF got 88% of points.
Give it a chance.

/Jarkko
Hi Jarkko,

I am testing SF_DEV_201015 just now. The difference with 15102104 must be small. I will post the results of the first half of this test -960 games- this morning. Good but not brilliant anyway.

Regards,

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

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

TESTING STOCKFISH DEVELOPMENT 201015 LAZY= 1920 GAMES
FIRST LEG. FIXED TIME CONTROL. 960 GAMES


Timestamp: 1445395741 Bench: 8855226

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

Stockfish 201015 64 POPCNT_X6 - Komodo 9.2 64-bit_x6 65.5 - 54.5 +30/=71/-19 54.58%
Stockfish 201015 64 POPCNT_X6 - Stockfish 190915 64 POP_6 61.0 - 59.0 +9/=104/-7 50.83%
Stockfish 201015 64 POPCNT_X6 - Houdini 4 x64_st_X6_CT0 75.0 - 45.0 +39/=72/-9 62.50%
Stockfish 201015 64 POPCNT_X6 - Gull 3 x64 XP 83.0 - 37.0 +49/=68/-3 69.17%

Score using 6 cores: = 284.5 – 195.5= 59.27%
480 Games: http://www.mediafire.com/download/enlloq0xcqew5sq

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
Stockfish 201015 64 BMI2_X8 - Komodo 9.2 64-bit_x8 61.5 - 58.5 +25/=73/-22 51.25%
Stockfish 201015 64 BMI2_X8 - Stockfish 190915 64 BMI2_8 61.5 - 58.5 +10/=103/-7 51.25%
Stockfish 201015 64 BMI2_X8 - Houdini 4 Pro x64_Ct0_8 78.0 - 42.0 +41/=74/-5 65.00%
Stockfish 201015 64 BMI2_X8 - Gull 3 x64 XPx8 80.5 - 39.5 +51/=59/-10 67.08%
Score using 8 real cores: = 281.5 – 198.5= 58.65%
480 Games: http://www.mediafire.com/download/vi11m99cwq1n50t

Against : Komodo 9.2 (3251) = 52.91% Stockfish 190915 (3262) = 51.04% Houdini 4 (3136) = 63.75%, Gull 3 XP (3103) = 68.12%

Average Elo of Oponents= 3.188

GLOBAL SCORE AFTER 960 GAMES AT FIXED TIME CONTROL = 566.0 – 394.0= 58.96%

ESTIMATED ELO PERFORMANCE FOR STOCKFISH DEVELOPMENT 201015 LAZY AFTER 960 GAMES AT FIXED TIME CONTROL= 3251


Error Bars = +/- 16.

At Fixed Time Control this is 13 Elo below the top scorer in my tests SF DEV 190915.

Regards,

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

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

TESTING STOCKFISH DEVELOPMENT LAZY 201015 = 1920 GAMES
SECOND LEG. INCREMENTAL TIME CONTROL. 960 GAMES


Timestamp: 1445395741 Bench: 8855226

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
Stockfish 201015 64 POPCNT_X6 - Komodo 9.2 64-bit_x6 61.5 - 58.5 +25/=73/-22 51.25%
Stockfish 201015 64 POPCNT_X6 - Stockfish 190915 64 POP 63.0 - 57.0 +10/=106/-4 52.50%
Stockfish 201015 64 POPCNT_X6 - Houdini 4 x64_st_X6_CT0 70.5 - 49.5 +36/=69/-15 58.75%
Stockfish 201015 64 POPCNT_X6 - Gull 3 x64 XP 85.5 - 34.5 +54/=63/-3 71.25%

Score using 6 cores: = 280.5 – 199.5= 58.44%

480 Games: http://www.mediafire.com/download/o9enyf688rsdpba

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
Stockfish 201015 64 BMI2_X8 - Komodo 9.2 64-bit_x8 57.0 - 63.0 +21/=72/-27 47.50%
Stockfish 201015 64 BMI2_X8 - Stockfish 190915 64 BMI2_8 58.0 - 62.0 +7/=102/-11 48.33%
Stockfish 201015 64 BMI2_X8 - Houdini 4 Pro x64_Ct0_8 77.5 - 42.5 +42/=71/-7 64.58%
Stockfish 201015 64 BMI2_X8 - Gull 3 x64 XPx8 90.0 - 30.0 +65/=50/-5 75.00%

Score using 8 real cores: = 282.5 – 197.5 = 58.85%

480 Games: http://www.mediafire.com/download/9hlk6y1dki76rhp

Against : Komodo 9.2 (3251) = 49.37% Stockfish 190915 (3262) = 50.41% Houdini 4 (3136) = 61.67%, Gull 3 XP (3103) = 73.12%

Average Elo of Oponents= 3.188

GLOBAL SCORE AFTER 960 GAMES AT INCREMENTAL TIME CONTROL= 563.0 – 397= 58.65%
ESTIMATED ELO FOR STOCKFISH DEV 201015 LAZY AT INCREMENTAL TIME CONTROL AFTER 960 GAMES= 3.249

================================

GLOBAL SCORE FOR STOCKFISH DEV 121015 AFTER 1.920 GAMES = 1.129.0 – 791.0= 58.80%

ESTIMATED GLOBAL ELO FOR STOCKFISH DEV 201015 LAZY AFTER 1.920 GAMES= 3.250


Error Bars +/- 11

This score is 12 Elo points below the Top Scorer so far in my tests Stockfish Dev. 190915 (3.262).

I noticed that this Stock DEV LAZY 201015 performs very well against the mentioned leader, but it is weaker against Houdini and Gull.

On the other hand I have not found bugs after 1.920 games, but I am not 100% sure that there is not any bug in these games.

Regards,

Tom.
Kohflote
Posts: 219
Joined: Wed Sep 19, 2007 11:07 am
Location: Singapore

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Kohflote »

Hi Tom,
I'm curious how good the last non-lazy version of SF. Would you like to test it, please?
Thank you,
Koh, Kah Huat
Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 9:09 pm

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

Kohflote wrote:Hi Tom,
I'm curious how good the last non-lazy version of SF. Would you like to test it, please?
Thank you,
Koh, Kah Huat
Hello my friend,

The latest Stockfish Development I tested before Lazy was 121015. It was also a 1.920 games test in the same conditions and against the same rivals. You can find the scores and games above in this page. Its Elo was exactly the same than SF Development Lazy 201015:

STOCKFISH DEVELOPMENT 121015= 3.250
STOCKFISH DEV 201015 LAZY = 3.250

So that I don't see a big impact -positive or negative- after the Lazy changes. Both are 12 ELO below the Top Scorer in my tests SF DEV 190915.

Regards,

Tom.
jarkkop
Posts: 198
Joined: Thu Mar 09, 2006 2:44 am
Location: Helsinki, Finland

Re: For Tom- or Maybe Anyone Who Knows!!

Post by jarkkop »

Thanks for your reply

Version stockfish_15103119_x64_modern
made a record of solving
r2qrb1k/1p1b2p1/p2ppn1p/8/3NP3/1BN5/PPP3QP/1K3RR1 w - - bm e5
at ply 29. So it could be a jump in performance.

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

Re: For Tom- or Maybe Anyone Who Knows!!

Post by Tomcass »

TESTING STOCKFISH DEVELOPMENT KP 021115 = 2.400 GAMES

This is a very fast compile by GarriCar13 (K.P.). This change has been made in line 392 of search.cpp :
rootDepth = Threads.main()->rootDepth + Depth(int(3.1 * log(1 + this->idx)));
This 3.1 digit works very well in my 6 and 8 cores computers.
Moreover Large Pages have been allowed

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

Stockfish 021115 KP_X6 - Komodo 9.2 64-bit_x6 132.0 - 108.0 +63/=138/-39 55.00%
Stockfish 021115 KP_X6 - Stockfish 190915 64 POPCNT_6 126.0 - 114.0 +26/=200/-14 52.50%
Stockfish 021115 KP_X6 - Houdini 4 Pro x64x9_Ct0 166.0 - 74.0 +104/=124/-12 69.17%
Stockfish 021115 KP_X6 - Gull 3 x64 XP 171.5 – 68.5 +112/=119/-9 71.46%

Score after 960 games using 6 cores: = 595.5 – 364.5= 62.03%
960 Games: http://www.mediafire.com/download/7j91qctk52e4xxm

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

Stockfish 021115 KP_X8 - Komodo 9.2 64-bit_x8 129.0 – 111.0 +50/=158/-32 53.75 %
Stockfish 021115 KP_X8 - Stockfish 190915 64 BMI2_8 123.5 - 116.5 +22/=203/-15 51.46%
Stockfish 021115 KP_X8 - Houdini 4 Pro x64_Ct0_8 169.5 - 70.5 +107/=125/-8 70.62%
Stockfish 021115 KP_X8 - Gull 3 x64 XPx8 168.5 - 71.5 +107/=123/-10 70.21%

Score after 960 games using 8 cores: 590.5 - 369.5 = 61.51%
http://www.mediafire.com/download/9v8k2es8467l814

Against : Komodo 9.2 (3251) = 54.37% Stockfish 190915 (3262) = 51.98% Houdini 4 (3136) = 69.89%, Gull 3 XP (3103) = 70.83%

Average Elo of Oponents= 3.188

GLOBAL SCORE FOR SF DEV 021115 KP AFTER 1920 GAMES AT FIXED TIME CONTROL= 1186.0 – 734.0 = 61.77%


ESTIMATED ELO FOR SF DEV 021115 KP AT FIXED TIME CONTROL AFTER 1920 GAMES= 3.270

This is the best performance ever in my tests at fixed time control. Only as a reference, it has performed 19 ELO better than STOCKFISH DEVELOPMENT 201015 LAZY.

Let’s see what happen at incremental. A 480 games test have just started.

Thank you very much, GarriCar13!

Regards,

Tom
beram
Posts: 1187
Joined: Wed Jan 06, 2010 3:11 pm

Re: For Tom- or Maybe Anyone Who Knows!!

Post by beram »

hi Tom,

Where can we get this 'garricar' SF DEV 021115 KP compile ?