SWCR-64: Houd, Fire, Loop M1-T, Equi, Cipo, CraftyNP

[Frank Quisinsky]

Hi there,

now after 8.166 of 11.040 games ...

ELO calculation (4, 6 updates added)
After 21.286 games (SWCR-64, all games)

Code: Select all

01. IPP Houdini 1.03a x64        2.952 33 32  391 79% 2.720 27% NEW
02. Rybka 4 x64                  2.940 19 18 1259 81% 2.691 28%
--. IPP Fire 1.31 x64            2.909 31 30  391 75% 2.722 35% NEW
--. Rybka 3 x64                  2.906 22 22  840 78% 2.688 28%
03. Stockfish 1.8.0 JA x64       2.904 21 21  900 76% 2.703 30%
--. Stockfish 1.7.1 JA x64       2.897 19 18 1120 76% 2.707 34%
04. Critter 0.80 x64             2.840 20 20  900 68% 2.706 33%
05. Naum 4.2 x64                 2.832 15 15 1499 68% 2.702 36%
--. Critter 0.70 x64             2.807 19 19  880 65% 2.702 38%
06. Shredder 12 w32              2.800 16 15 1459 64% 2.700 35%
06. Komodo 1.2 JA x64            2.800 16 16 1260 65% 2.695 39%
--. Komodo 1.0 JA x64            2.792 20 20  840 64% 2.694 40%
--. Shredder 12 x64              2.789 18 18 1080 61% 2.708 35%
08. Spark 0.5 x64                2.740 16 16 1259 56% 2.697 38%
09. Thinker 5.4d Inert x64       2.735 15 15 1500 54% 2.704 39%
10. Zappa Mexico II x64          2.724 15 15 1500 53% 2.704 39%
--. Spark 0.4 x64                2.718 19 19  840 53% 2.697 40%
11. Protector 1.3.4 JA x64       2.716 15 15 1459 52% 2.705 36%
12. Fruit 09_07_05 x64           2.710 15 15 1500 51% 2.705 34%
--. Critter 0.60 x64             2.701 20 20  840 50% 2.698 38%
13. Hannibal 1.0a x64            2.692 19 19  900 47% 2.713 36%
14. Sjeng WC-2008 x64            2.687 15 15 1499 48% 2.706 36%
--. Protector 1.3.5 x64          2.685 19 19  840 47% 2.707 39%
15. Junior 11.2 x64              2.683 17 16 1260 48% 2.699 31%
16. Onno 1.2.70 x64              2.676 16 16 1259 47% 2.699 37%
--. Onno 1.1.1 x64               2.672 19 19  840 46% 2.700 40%
--. Junior 11.1a x64             2.649 20 20  840 43% 2.701 32%
17. Loop 2007 x64                2.632 15 16 1400 40% 2.707 37%
--. Loop M1-T x64                2.622 30 30  372 35% 2.735 34% -10 ELO
18. Equinox 0.83 x64             2.619 30 31  383 35% 2.734 31% NEW
--. Twisted Logic 20100131x x64  2.612 18 18 1120 35% 2.717 32%
19. Umko 1.0 x64                 2.607 19 20  900 35% 2.717 36% (ponder not possible)
20. SmarThink 1.20 x64           2.604 15 15 1499 36% 2.708 34%
21. Crafty 23.3 JA x64           2.591 20 20  900 33% 2.717 33%
--. Cipollino 3.25 x64           2.588 31 31  372 30% 2.736 33% NEW
22. BugChess2 1.7 x64            2.564 21 22  800 29% 2.720 33%
23. Scorpio 2.6 JA x64           2.556 18 18 1120 28% 2.719 32%
--. Crafty 23.2 JA x64           2.556 18 18 1120 28% 2.719 30%
24. Chronos 1.99 x64             2.553 18 18 1120 27% 2.719 33% (ponder not possible)
--. Crafty 23.3 JA x64 NP        2.547 31 32  391 25% 2.738 30% (without ponder = -44 ELO)
25. Daydreamer 1.75 JA x64       2.522 19 19 1120 24% 2.720 30%
26. Tornado 3.6.7 x64            2.480 23 24  800 19% 2.724 24%

LIVE tournament table available under ...

Frank's Chess Page, SWCR
http://www.amateurschach.de

Best
Frank

[Sedat Canbaz]

Hello Frank,

Code: Select all

--. Crafty 23.3 JA x64 NP        2.547 31 32  391 25% 2.738 30% (without ponder = -44 ELO)

Wow...44 elo -really a lot of difference

Btw,currently i am testing Crafty 23.3 JA x64 T4 with ponder off at:
http://sedatchess.110mb.com/index.php?p=1_58

And maybe later i hope to test Crafty 23.3 JA x64 T4 with ponder on in my SCCT (10 Min Game) Rating List and then we can compare the elo difference,even we can see the influence of using 4 threads plus pondering under different conditions


Best,
Sedat

[Frank Quisinsky]

Hi Sedat,

yes, this is interesting ...
I will follow your test with Crafty ... ponder on/off.

In the past I test it one different times. Averagely it should be 27 ELO difference. I think with more games Crafty 23.3 JA x64 NP will win points. After around 600 games is more as 35 ELO really a little sensation.

So we have to wait ...
Monday next week I have around 700 games ...
Let us look Monday of the Crafty 23.3 JA x64 NP results.

Best
Frank

[Frank Quisinsky]

Hi there,

at the moment Houdini 1.03a x64 isn't running. I can't manage tournament crashes in the next 4 days. Houdini crasht in the first 460 on 15 times. In these days the most tournaments will stopped if Houdini crasht, so I manage a tournament stop with the "x-funktion" in Shredder Classic 4 GUI *.sto file. The Houdini games will be played Monday.

460 of 920 games are played now for the new 6 engines.
Next ELO calculation will be available next week Monday (very late, German time).

Have fun with SWCR ...

Code: Select all

ELO calculation (5)
After 21.715 games (SWCR-64, all games)

01. IPP Houdini 1.03a x64        2.952 31 30  461 80% 2.719 27% NEW
02. Rybka 4 x64                  2.939 19 18 1280 81% 2.690 28%
--. Rybka 3 x64                  2.905 22 22  840 78% 2.688 28%
03. Stockfish 1.8.0 JA x64       2.904 21 21  920 76% 2.702 30%
--. IPP Fire 1.31 x64            2.897 28 28  461 74% 2.721 36% NEW
--. Stockfish 1.7.1 JA x64       2.896 19 18 1120 76% 2.706 34%
04. Critter 0.80 x64             2.838 20 20  921 68% 2.705 33%
05. Naum 4.2 x64                 2.831 15 15 1520 68% 2.701 35%
--. Critter 0.70 x64             2.807 19 19  880 65% 2.701 38%
06. Shredder 12                  2.800 15 15 1480 64% 2.699 35%
07. Komodo 1.2 JA x64            2.799 16 16 1281 65% 2.694 39%
--. Komodo 1.0 JA x64            2.791 20 20  840 64% 2.693 40%
--. Shredder 12 x64              2.788 18 17 1080 61% 2.707 35%
08. Spark 0.5 x64                2.740 16 16 1280 56% 2.696 38%
09. Thinker 5.4d Inert x64       2.735 15 15 1521 54% 2.703 40%
10. Zappa Mexico II x64          2.723 15 15 1521 53% 2.703 39%
--. Spark 0.4 x64                2.718 19 19  840 53% 2.697 40%
11. Protector 1.3.4 JA x64       2.715 15 15 1480 52% 2.704 36%
12. Fruit 09_07_05 x64           2.709 15 15 1521 51% 2.704 34%
--. Critter 0.60 x64             2.700 20 20  840 50% 2.697 38%
13. Hannibal 1.0a x64            2.690 19 19  921 47% 2.711 36%
14. Sjeng WC-2008 x64            2.685 15 15 1520 48% 2.705 36%
--. Protector 1.3.5 x64          2.685 19 19  840 47% 2.707 39%
15. Junior 11.2 x64              2.682 16 16 1282 48% 2.698 31%
16. Onno 1.2.70 x64              2.676 16 16 1280 47% 2.698 38%
--. Onno 1.1.1 x64               2.671 19 19  840 46% 2.699 40%
--. Junior 11.1a x64             2.648 20 20  840 43% 2.700 32%
17. Loop M1-T x64                2.634 27 27  461 37% 2.732 35% + 2 ELO
--. Loop 2007 x64                2.632 15 16 1400 40% 2.706 37%
--. Twisted Logic 20100131x x64  2.611 18 18 1120 35% 2.716 32%
18. Equinox 0.83 x64             2.607 28 28  461 33% 2.734 30% NEW
19. Umko 1.0 x64                 2.606 19 19  921 35% 2.715 36% (ponder not possible)
20. SmarThink 1.20 x64           2.604 15 15 1520 36% 2.707 34%
21. Crafty 23.3 JA x64           2.592 19 20  921 33% 2.716 34%
--. Cipollino 3.25 x64           2.581 28 28  461 29% 2.735 32% NEW
22. BugChess2 1.7 x64            2.564 21 22  800 29% 2.719 33%
23. Scorpio 2.6 JA x64           2.555 18 18 1120 28% 2.718 32%
--. Crafty 23.2 JA x64           2.555 18 18 1120 28% 2.718 30%
24. Chronos 1.99 x64             2.552 18 18 1120 27% 2.719 33% (ponder not possible)
--. Crafty 23.3 JA x64 NP        2.550 28 29  475 26% 2.736 31% (without ponder = -42 ELO)
25. Daydreamer 1.75 JA x64       2.521 19 19 1120 24% 2.720 30%
26. Tornado 3.6.7 x64            2.479 23 24  800 19% 2.723 24%

Frank's Chess Page, SWCR
http://www.amateurschach.de

Best
Frank

[Frank Quisinsky]

Hi there,

now the 6 new engines played 736 / 920 games.
Here the actual rating list from SWCR-64:

ELO calculation (7)
After 23.176 games (SWCR-64, all games)

Code: Select all

01. IPP Houdini 1.03a x64        2.949 24 24  736 79% 2.718 28% NEW
02. Rybka 4 x64                  2.938 18 18 1352 81% 2.690 28%
--. Rybka 3 x64                  2.905 22 22  840 78% 2.687 28%
03. Stockfish 1.8.0 JA x64       2.904 20 20  992 76% 2.701 31%
--. IPP Fire 1.31 x64            2.901 23 22  736 75% 2.720 36% NEW
--. Stockfish 1.7.1 JA x64       2.896 19 18 1120 76% 2.706 34%
04. Critter 0.80 x64             2.838 19 19  992 68% 2.704 33%
05. Naum 4.2 x64                 2.831 15 15 1592 68% 2.700 35%
--. Critter 0.70 x64             2.806 20 19  880 65% 2.701 38%
06. Komodo 1.2 JA x64            2.801 16 16 1352 65% 2.694 39%
07. Shredder 12 w32              2.800 15 15 1552 64% 2.698 35%
--. Komodo 1.0 JA x64            2.791 20 20  840 64% 2.693 40%
--. Shredder 12 x64              2.788 18 18 1080 61% 2.707 35%
08. Spark 0.5 x64                2.740 16 16 1352 56% 2.696 38%
09. Thinker 5.4d Inert x64       2.735 14 14 1592 55% 2.702 39%
10. Zappa Mexico II x64          2.723 14 14 1592 53% 2.703 39%
--. Spark 0.4 x64                2.717 19 19  840 53% 2.696 40%
11. Protector 1.3.4 JA x64       2.714 15 15 1552 52% 2.703 37%
12. Fruit 09_07_05 x64           2.706 15 15 1592 50% 2.703 34%
--. Critter 0.60 x64             2.700 19 19  840 50% 2.697 38%
13. Hannibal 1.0a x64            2.691 18 18  992 48% 2.710 36%
14. Sjeng WC-2008 x64            2.687 15 14 1592 48% 2.704 35%
--. Protector 1.3.5 x64          2.684 19 19  840 47% 2.706 39%
15. Junior 11.2 x64              2.682 16 16 1352 48% 2.697 30%
16. Onno 1.2.70 x64              2.675 16 16 1352 47% 2.698 38%
--. Onno 1.1.1 x64               2.671 19 19  840 46% 2.698 40%
--. Junior 11.1a x64             2.648 20 20  840 43% 2.700 32%
17. Loop 2007 x64                2.632 15 16 1400 40% 2.706 37%
--. Loop M1-T x64                2.625 21 22  736 35% 2.732 35% - 7 ELO
--. Twisted Logic 20100131x x64  2.611 18 18 1120 35% 2.716 32%
18. Umko 1.0 x64                 2.606 19 19  992 35% 2.713 36% (ponder not possible)
19. SmarThink 1.20 x64           2.602 15 15 1592 36% 2.706 34%
20. Equinox 0.83 x64             2.596 22 22  736 32% 2.733 32% NEW
21. Crafty 23.3 JA x64           2.592 19 19  992 34% 2.714 34%
--. Cipollino 3.25 x64           2.571 22 23  736 28% 2.734 32% NEW
22. BugChess2 1.7 x64            2.564 21 21  800 29% 2.719 33%
23. Scorpio 2.6 JA x64           2.555 18 18 1120 28% 2.718 32%
--. Crafty 23.2 JA x64           2.555 18 18 1120 28% 2.718 30%
24. Chronos 1.99 x64             2.552 18 18 1120 27% 2.718 33% (ponder not possible)
--. Crafty 23.3 JA x64 NP        2.546 23 23  736 25% 2.735 31% (without ponder = -46 ELO)
25. Daydreamer 1.75 JA x64       2.521 19 19 1120 24% 2.719 30%
26. Tornado 3.6.7 x64            2.479 23 24  800 19% 2.723 24%

The final results are Sunday in the evening available.

More information can be found under:

Frank's Chess Page, SWCR
http://www.amateurschach.de

Best
Frank

[Frank Quisinsky]

Hi there,

under 500 games are only to play.
Latest calculation before I create all the updates (games and so one) Sunday on me webpage.

Code: Select all

ELO calculation (8)
After 23.714 games (SWCR-64, all games)

01. IPP Houdini 1.03a x64        2.948 22 21  920 79% 2.718 29% NEW
02. Rybka 4 x64                  2.937 18 18 1375 81% 2.691 28%
--. Rybka 3 x64                  2.905 22 22  840 78% 2.687 28%
03. Stockfish 1.8.0 JA x64       2.904 20 20 1016 76% 2.702 31%
--. IPP Fire 1.31 x64            2.900 22 21  811 74% 2.721 36% NEW
--. Stockfish 1.7.1 JA x64       2.896 19 18 1120 76% 2.706 34%
04. Critter 0.80 x64             2.838 19 19 1016 68% 2.705 33%
05. Naum 4.2 x64                 2.831 15 15 1616 68% 2.701 35%
--. Critter 0.70 x64             2.806 20 19  880 65% 2.701 38%
06. Komodo 1.2 JA x64            2.800 16 16 1376 65% 2.695 39%
06. Shredder 12 w32              2.800 15 15 1575 64% 2.699 35%
--. Komodo 1.0 JA x64            2.791 20 20  840 64% 2.693 40%
--. Shredder 12 x64              2.788 18 18 1080 61% 2.707 35%
08. Spark 0.5 x64                2.740 16 16 1375 56% 2.697 38%
09. Thinker 5.4d Inert x64       2.736 14 14 1616 55% 2.703 39%
10. Zappa Mexico II x64          2.723 14 14 1616 53% 2.703 39%
--. Spark 0.4 x64                2.717 19 19  840 53% 2.696 40%
11. Protector 1.3.4 JA x64       2.714 15 15 1575 52% 2.704 37%
12. Fruit 09_07_05 x64           2.706 15 14 1616 50% 2.704 34%
--. Critter 0.60 x64             2.700 19 19  840 50% 2.697 38%
13. Hannibal 1.0a x64            2.689 18 18 1016 47% 2.710 35%
14. Sjeng WC-2008 x64            2.687 15 14 1615 48% 2.704 35%
--. Protector 1.3.5 x64          2.684 19 19  840 47% 2.706 39%
15. Junior 11.2 x64              2.681 16 16 1375 48% 2.698 30%
16. Onno 1.2.70 x64              2.678 16 16 1375 47% 2.699 38%
--. Onno 1.1.1 x64               2.671 19 19  840 46% 2.698 40%
--. Junior 11.1a x64             2.648 20 20  840 43% 2.700 32%
17. Loop 2007 x64                2.631 15 15 1440 40% 2.703 37%
--. Loop M1-T x64                2.627 20 21  811 36% 2.733 35% - 4 ELO
--. Twisted Logic 20100131x x64  2.611 18 18 1120 35% 2.716 32%
18. Umko 1.0 x64                 2.608 18 18 1016 35% 2.714 37% (ponder not possible)
19. SmarThink 1.20 x64           2.602 15 15 1615 36% 2.706 34%
20. Equinox 0.83 x64             2.598 20 21  851 32% 2.729 33% NEW
21. Crafty 23.3 JA x64           2.593 18 19 1016 34% 2.714 34%
--. Cipollino 3.25 x64           2.569 21 22  811 28% 2.735 30% NEW
22. BugChess2 1.7 x64            2.564 21 21  800 29% 2.719 33%
23. Scorpio 2.6 JA x64           2.555 18 18 1120 28% 2.718 32%
--. Crafty 23.2 JA x64           2.555 18 18 1120 28% 2.718 30%
24. Chronos 1.99 x64             2.552 18 18 1120 27% 2.718 33% (ponder not possible)
--. Crafty 23.3 JA x64 NP        2.544 21 22  824 25% 2.736 30% (without ponder = -49 ELO)
25. Daydreamer 1.75 JA x64       2.521 19 19 1120 24% 2.719 30%
26. Tornado 3.6.7 x64            2.479 23 24  800 19% 2.723 24%

Best
Frank

[bob]

Hi Sedat,

for the next comming tournament I am thinking about a second Crafty test. In SWCR all engines playing with all the different available 4-pieces TBs. In the past I find out that engines with 5-pieces TBs lost 15-20 ELO in eng-eng tournaments.

So, I can test it again with WB Crafty 23.3 JA x64.
Means after the still running tournament I can add Crafty 23.3 JA x64 5Tbs.

What do you think?
Do you think it isn't necessary to do it again, or interesting to do it again?

Best
Frank

Interesting, but perhaps a waste of time. I just did a 30,000 game test for each of 3 versions of Crafty. One with no egtbs at all, one using them all and probing normally, one using them all, but probing less aggressively. No Elo difference that was measurable. After 30,000 games by each version, they were all within 2 Elo of each other, with an error bar of +/-3...

[Frank Quisinsky]

Hi Bob,

yes, think so!

Do you saw the results from Crafty 23.3 JA x64 NP (NP for "No Ponder") in the ponder on tourney. 2 Crafty`s are playing, one version with ponder, one version without ponder. The different is at the moment 47 ELO after around 900 games with 40 moves in 10 minutes. I never expects 47 ELO, I expecting 25-30 ELO. A little sensation for myself.

BTW: After around 900 games no "Lost on time games" for the NON ponder version.

Best
Frank

[bob]

Hi Bob,

yes, think so!

Do you saw the results from Crafty 23.3 JA x64 NP (NP for "No Ponder") in the ponder on tourney. 2 Crafty`s are playing, one version with ponder, one version without ponder. The different is at the moment 47 ELO after around 900 games with 40 moves in 10 minutes. I never expects 47 ELO, I expecting 25-30 ELO. A little sensation for myself.

BTW: After around 900 games no "Lost on time games" for the NON ponder version.

Best
Frank

The 47 makes sense to me. Recent tests suggest +100 Elo for doubling processor speed. Pondering is somewhere in the 50% range overall (ponder hit rate is always higher, but a ponder hit is not always an instant move). 50% more time (roughly) could scale to 50% of that 100 Elo improvement (roughly)...

[Frank Quisinsky]

Hi Bob,

for some years I tested it on two times. I had in the first case 300 games only and in the second 450 games. 24 ELO and 29 ELO, so I expected around 27 ELO.

Its very interesting for myself.
I think a tourney with ponder = on is much more interesting as a tourney with ponder = off and perhaps 2 cores.

Thanks for your comments, but after all my test results from the past it is a little sensation for myself.

You can be sure, if you wrote before, that it is 50 ELO ... my comment will be ... never ever!

So we all learn daily and this after so many years computer chess.

Best
Frank