The great 40/120 Tournament

[krazyken]

Finished with Bikjump! It finishes a Little ahead of Alf.

Code: Select all

1: The Baron 2.23                     92.5 / 132
2: amundsen 0.70                      57.5 / 128
3: Beowulf v2.4                       49.0 / 132
4: BikJump v2.01 (64-bit)             43.5 / 130
5: Alf v1.09                          42.0 / 132

Bikjump has some bugs in endgame lookup, getting losses in drawn endings. It is also seriously sloppy with king safety. But it can get a surprisingly strong game from time to time. Amundsen 0.75 is being added in, it still seems to have problems with draws and executing a mate in 1. Toga is being pulled out because that build wasn't stable, I'm running some tests this week to find the best build to include. Mediocre 0.334 is starting stronger than 0.332 did.

Code: Select all

   1 Rybka 2.2n2 mp                     2660  334  334    10  100%  2155    0% 
   2 Fruit 2.3.1                        2660  334  334    10  100%  2155    0% 
   3 HIARCS 12.1 MP                     2548  254  254    10   95%  2155   10% 
   4 Gambit Fruit 1.0 Beta 4bx          2548  254  254    10   95%  2155   10% 
   5 Scorpio 2.02                       2548  254  254    10   95%  2155   10% 
   6 Twisted Logic 20080620             2548  254  254    10   95%  2155   10% 
   7 Kiwi 0.6d                          2542  254  254    10   95%  2155   10% 
   8 HIARCS 11.2 MP                     2542  254  254    10   95%  2155   10% 
   9 Glaurung 2.1                       2542  254  254    10   95%  2155   10% 
  10 Sloppy-0.2.0                       2510  261  261    10   90%  2155    0% 
  11 Thinker53B                         2499  256  256    10   90%  2155    0% 
  12 Deep Shredder 11 UCI               2496  252  252    10   90%  2155    0% 
  13 bright-0.3a                        2460  219  219    10   90%  2155   20% 
  14 Spike 1.2 Turin                    2456  220  220    10   90%  2155   20% 
  15 Shredder 9.14 UCI                  2456  220  220    10   90%  2155   20% 
  16 Crafty-22.8                        2441  224  224    10   85%  2155   10% 
  17 WildCat 8                          2430  216  216    10   85%  2155   10% 
  18 Naum 2.0                           2417  217  217    10   85%  2155   10% 
  19 Hamsters 0.7.1                     2413  221  221    10   85%  2155   10% 
  20 Glaurung-lodz                      2413  198  198    10   85%  2155   30% 
  21 The Baron 2.23                     2402   56   56   132   70%  2197   22% 
  22 HIARCS 10 UCI                      2387  215  215    10   80%  2155    0% 
  23 Booot 4.14.0                       2377  198  198    10   80%  2155   20% 
  24 Deep Sjeng 1.6                     2370  199  199    10   80%  2155   20% 
  25 Colossus 2008b                     2341  188  188    10   75%  2155   30% 
  26 Homer 2.01 UCI                     2318  183  183    10   75%  2155   30% 
  27 typhoon 1.00                       2311  206  206    10   70%  2155    0% 
  28 Pepito v1.59 (Conservador)         2306  194  194    10   70%  2155   20% 
  29 Natwarlal v0.14                    2302  188  188    10   70%  2155   20% 
  30 resp                               2301  190  190    10   70%  2155   20% 
  31 Diablo 0.5.1                       2300  188  188    10   70%  2155   20% 
  32 ProDeo 1.6                         2299  185  185    10   75%  2155   30% 
  33 Joker 1.1.14f                      2298  187  187    10   70%  2155   20% 
  34 amundsen 0.75                      2291  170  170    10   70%  2165   40% 
  35 Gaia 3.5                           2274  184  184    10   70%  2155   20% 
  36 phalanx                            2253  180  180    10   65%  2155   30% 
  37 GreKo 6.1                          2233  167  167    10   65%  2155   50% 
  38 Mediocre v0.334                    2218  173  173    10   55%  2179   30% 
  39 amundsen 0.70                      2167   59   59   128   45%  2199   12% 
  40 romi                               2151  177  177    10   50%  2155   20% 
  41 Gibbon 2.42c                       2135  170  170    10   50%  2155   40% 
  42 Flux 2.2                           2122  187  187    10   45%  2155   10% 
  43 ZCT0.3.2483                        2115  165  165    10   45%  2155   50% 
  44 Beowulf v2.4                       2103   58   58   132   37%  2201   12% 
  45 Witz Alpha 21                      2092  171  171    10   40%  2155   40% 
  46 gnuchess                           2084  167  167    10   40%  2155   40% 
  47 RattateChess 1.0-rc1 Bologna 2007  2078  197  197    10   40%  2155    0% 
  48 BikJump v2.01 (64-bit)             2070   60   60   130   33%  2205   16% 
  49 Mediocre v0.332                    2036  212  212     8   31%  2176   13% 
  50 Alf v1.09                          2030   60   60   132   32%  2203   17% 
  51 OliThink 5.1.8alpha                1970  202  202    10   25%  2155   10% 
  52 KnightCap                          1954  179  179    10   20%  2155   40% 
  53 King's Out 0.2.41                  1932  194  194    10   20%  2155   20% 
  54 JChess.bat                         1904  191  191    10   15%  2155   30% 
  55 colchess                           1881  214  214    10   15%  2155   10% 
  56 Gullydeckel 2.16.pl1               1851  211  211    10   10%  2155   20% 
  57 TSCP                               1851  211  211    10   10%  2155   20% 
  58 Roce 0.0380                        1850  211  211    10   10%  2155   20% 
  59 Eden 0.0.12                        1779  245  245    10    5%  2155   10% 
  60 HoiChess 0.10.3                    1777  245  245    10    5%  2155   10% 
  61 Xpdnt 080927                       1777  245  245    10    5%  2155   10% 
  62 El Turco 0.93                      1663  323  323    10    0%  2155    0% 
  63 hanzo                              1663  323  323    10    0%  2155    0% 
  64 Xadreco v.5.7, by Beco             1663  323  323    10    0%  2155    0% 
  65 EXchess_console                    1663  323  323    10    0%  2155    0% 
  66 Gray Matter rev 1522M              1663  323  323    10    0%  2155    0% 
  67 mscp                               1663  323  323    10    0%  2155    0% 

Will be adding Arasan and updating Glaurung shortly. Possibly Grapefruit and Cyclone may be joining as well. Games can be found here, and extra games from engines that have been replaced or pulled can be found here.

[krazyken]

Samples from Bikjump:

shortest win

[Event "Computer Chess Game"]
[Site "Mulert.local"]
[Date "2008.12.14"]
[Round "1"]
[White "BikJump v2.01 (64-bit)"]
[Black "TSCP"]
[Result "1-0"]
[TimeControl "40/7200"]

1. d4 e6 2. e4 d5 {-0.26/8 3:25} 3. Nc3 Bb4 {+0.19/8 56} 4. e5
Bxc3+ {+0.32/8 1:20} 5. bxc3 Nc6 {+0.11/8 55} 6. Bd3 {+0.36/16 1:17}
Nge7 {+0.01/8 1:40} 7. Nf3 {+0.31/16 48} Bd7 {-0.02/8 2:02} 8.
Rb1 {+0.41/16 1:49} Rb8 {-0.07/8 2:28} 9. Ba3 {+0.36/15 30}
O-O {+0.00/8 1:16} 10. Bxh7+ {+1.09/15 1:53} Kxh7 {+0.28/7 48} 11.
Ng5+ {+1.09/15 2:59} Kh6 {-0.08/7 1:35} 12. Qg4 {+5.08/14 1:36}
Nxe5 {+0.63/7 39} 13. Qh4+ {+99.87/7 1} Kg6 {-7.89/7 47} 14.
Qh7+ {+99.89/6 0} Kf6 {-99.90/7 2:08} 15. dxe5+ {+99.91/5 0}
Kxe5 {-99.92/6 20} 16. Nf3+ {+99.93/4 0} Kf6 {-99.94/5 1} 17.
Qh5 {+99.95/3 0} e5 {-99.96/4 0} 18. Qg5+ {+99.97/1 0} Ke6 {-99.98/2 0} 19.
Qxe5# {+99.99/1 0}
{White mates} 1-0

shortest loss

[Event "Computer Chess Game"]
[Site "Mulert.local"]
[Date "2008.12.14"]
[Round "2"]
[White "Naum 2.0"]
[Black "BikJump v2.01 (64-bit)"]
[Result "1-0"]
[TimeControl "40/7200"]

1. d4 Nf6 2. c4 g6 3. Nf3 Bg7 4. Nc3 O-O 5. e4 d6 6. Be2 Bg4 7. Be3 Nfd7 8.
Rc1 e5 9. d5 f5 10. exf5 {+0.50/17 2:41} Bxf5 {-0.30/14 1:39} 11.
Ng5 {+1.02/17 52} Nc5 {-0.66/15 1:48} 12. b4 {+1.80/19 2:18}
h6 {-0.93/15 1:08} 13. bxc5 {+1.73/19 2:27} hxg5 {-1.06/16 1:50} 14.
cxd6 {+2.35/18 1:59} cxd6 {-1.50/18 8:24} 15. g4 {+2.09/19 3:15}
Bc8 {-1.50/16 0} 16. Qc2 {+2.30/17 47} Kh7 {-1.63/16 3:24} 17.
Ne4 {+2.23/17 2:11} Bf6 {-1.92/16 8:06} 18. c5 {+2.85/17 3:35}
dxc5 {-1.97/15 3:15} 19. h4 {+3.84/17 8:20} gxh4 {-1.38/14 1:30} 20.
d6 {+3.84/16 0} Qa5+ {-1.19/12 1:59} 21. Bd2 {+5.45/17 3:37}
Qb6 {-1.19/11 0} 22. g5 {+7.25/13 1:19} Bd8 {-1.21/13 1:03} 23.
Qc3 {+13.58/13 1:59} Nc6 {-4.93/12 43} 24. Qg3 {+13.99/13 3:25}
h3 {-4.93/10 0} 25. Qh4+ {+99.85/6 0} Kg8 {-99.86/9 1} 26.
Bc4+ {+99.87/5 0} Rf7 {-99.88/8 0} 27. Qh6 {+99.87/4 0} Be6 {-99.90/6 0}
28. Bxe6 {+99.91/3 0} Qb3 {-99.90/3 0} 29. axb3 {+99.93/3 0}
Bxg5 {-99.92/1 0} 30. Nxg5 {+99.95/3 0} Nd8 {-99.94/1 0} 31.
Qh7+ {+99.97/3 0} Kf8 {-99.98/1 0} 32. Qh8# {+99.99/3 0}
{Computer wins} 1-0

[abik]

Kenny,
I appreciate your tests and also the detailed reports you gave me through PM. I hope to get some spare time again soon to look at those Nalimov endgame lookup bugs and, thus, avoid giving away won endgames. Thanks again!
Aart

[Zlaire]

It's fascinating how wildly different Mediocre performs in different tests. And for some reason I usually get much better results when running the tests myself.

The only real difference between v0.332 and v0.334 is that Mediocre supports polyglot opening books now.

Next version should be better though, but it's a couple of weeks away (going out of town for a while now).

[Graham Banks]

It's fascinating how wildly different Mediocre performs in different tests. And for some reason I usually get much better results when running the tests myself.

The only real difference between v0.332 and v0.334 is that Mediocre supports polyglot opening books now.

Next version should be better though, but it's a couple of weeks away (going out of town for a while now).

Hi Jonatan,

does this mean that the results of Mediocre 0.332 and 0.334 can be combined or should we still treat them as separate engines for rating list purposes?

Cheers, Graham.

[Zlaire]

I'd say combine them.

There might be some off-chance that the cleanup of the code (mainly rewriting comments etc.) made some unintended difference but most likely not.

[Graham Banks]

I'd say combine them.

There might be some off-chance that the cleanup of the code (mainly rewriting comments etc.) made some unintended difference but most likely not.

Thanks Jonatan. Hope you had a nice Christmas.

Regards, Graham.

[Michel]

Toga is being pulled out because that build wasn't stable,

I noticed that some versions of Toga on some compilers do not work well with some optimization settings. It would be interesting to understand this.

[krazyken]

Toga is being pulled out because that build wasn't stable,

I noticed that some versions of Toga on some compilers do not work well with some optimization settings. It would be interesting to understand this.

Yes, it has come a long way from it's stable Fruit roots. It appears to be connected to the threading code, as reducing the number of threads increases the time between failure. I will try less aggressive optimizations to see where I get. Have you run many longer time control games with many threaded Toga?

P.S. The versions of DLT and Grapefruit on your site do not run at all with multiple threads on My mac.

[swami]

Are they all really MAC compatible, Ken?

Perhaps you'd love to operate for Hamsters at CCT? See if you can get the Ok from Stefan for operator - Shredder.