CENT - Sub Pool started

[Rebel]

CENT is a computer-computer tournament starting from (mostly strategic) endgame positions only. The goal of CENT is to measure the end game playing strength in comparison with the CCRL and CEGT rating lists, if the order is going to fluctuate and with how much. For engine authors it may serve as an inspiration to improve in this area.

...

http://rebel13.nl/misc/cent.html

-----------

Sub Pool is started first. With its 4 participants (only) producing just 252 games it will be finished in a couple of days. Thereafter the Main Pool will start, approx. duration 20-24 days.

Aside from the official live support there is a more neutral way of following the games.

http://www.top-5000.nl/pgn4web/live1.html

It's my intention to make this freeware.

MATCH is a web utility to broadcast computer engine-engine matches live on the internet. With MATCH you can broadcast up to 4 comp-comp matches at same time live on the internet. Owners with a webspace plus a text editor and a minimum of HTML knowledge can setup such live matches in a relative short time. MATCH will work with interfaces that support the following 2 features as shown in the screenshot of Arena (Options -> Appearance -> PGN). More...

[Rebel]

Winner of the Sub-Pool is WaDuuttie, congrats Maarten.

Code: Select all

No. Engine                      1     2     3     4     Score     [%]
-----------------------------------------------------------------------
 1 WaDuuttie                 xxxx  21.5  22.5  31.0   75.0 / 126 (59.5%)
 2 Texel 1.06a45             20.5  xxxx  21.5  31.0   73.0 / 126 (57.9%)
 3 Hannibal1.5x64            19.5  20.5  xxxx  30.0   70.0 / 126 (55.5%)
 4 Zurichess                 11.0  11.0  12.0  xxxx   34.0 / 126 (26.9%)

http://rebel13.nl/misc/cent.html#subpool

[Frank Quisinsky]

Hi Ed,

I don't need CEGT and CCRL (have more games with 1 Core as both togehter with 1 Core and TOP-57) ...

Have it ...
Easy to create with a database each one against each other 50 games ...
Here the Rating List from move number 100-299 ... with around 6% errorbar (means games with move number before 100 ... games endgames started much earlier). But enough games I have and the error margin have all 55 engines. So the ratings are good (not very good, but good).

You can see how strong Tornado, Scorpio or Bobcat and EXchess (different engines plays a fantastic endgame) are in endgames comparing to the other playing phases.

Have fun with your new idea.
I am watching the results (like it to compare a bit with my own work).

Interesting is:
In endgames the Elo range between the TOP-55 is ...
ONLY 244 Elo !!!!!

Code: Select all

 FCP Rating List
  Each one against each other!
  ----------------------------

  Date           : May 12th, 2016 (05:00)
  Games          : 77.000 / 20.470 = 26.584%
  Move Number    : 100 - 299
  Version        : 4.10
  Elo average    : 2816
  Elo range      : 244.41
  Conditions     : http://www.amateurschach.de/main/_conditions.htm

  Generated with : Ordo 1.2.6 x64 (Ratings for chess and other games)
  Ordo_126-x64.exe -a 2816 -p 299.pgn -o ordo126.txt -b columns.txt -N2 -QEWDV -U "0,1,4,5,7,8,9,3,10,2,11,12,14" -s1000
  
  https://sites.google.com/site/gaviotachessengine/ordo
  by Miguel A. Ballicora (Argentina)



   # Player                           :      Elo  Games  Score%  won  draw  lost  Points  Draw%  Error   OppAvg   OppE   OppD
  ---------------------------------------------------------------------------------------------------------------------------
   1 Komodo 9.42 x64                  :  2976.77    448    67.9  175   258    15   304.0   57.6  20.44  2840.17  14.47   46.2
   2 Stockfish 24Apr2016 BMI2 x64     :  2966.08    541    67.0  186   353     2   362.5   65.2  18.17  2835.89  14.53   49.3
   3 Houdini 4 STD B x64              :  2952.02    803    65.9  289   480    34   529.0   59.8  14.77  2832.32  14.55   50.0
   4 Fire 4 x64                       :  2928.00    686    63.2  213   441    32   433.5   64.3  16.24  2829.42  14.44   50.7
   5 GullChess 3.0 BMI2 x64           :  2912.68    638    61.4  178   428    32   392.0   67.1  16.31  2828.21  14.39   50.1
   6 Fritz 15 x64                     :  2890.77    683    58.6  166   469    48   400.5   68.7  15.44  2828.04  14.39   50.9
   7 Equinox 3.30 x64                 :  2888.63    697    58.6  176   465    56   408.5   66.7  14.97  2825.14  14.40   51.9
   8 Nirvanachess 2.2 POP x64         :  2886.17    774    58.6  217   473    84   453.5   61.1  14.73  2823.87  14.29   52.0
   9 Andscacs 0.86 POP x64            :  2881.81    647    57.4  148   447    52   371.5   69.1  15.64  2828.01  14.39   50.4
  10 Protector 1.9.0 x64              :  2880.16    697    57.7  170   464    63   402.0   66.6  14.74  2824.38  14.35   52.0
  11 Hannibal 1.5 x64                 :  2876.12    872    58.2  248   519   105   507.5   59.5  13.22  2816.97  14.27   53.2
  12 Critter 1.6a x64                 :  2870.87    714    56.0  144   512    58   400.0   71.7  14.37  2826.82  14.41   51.7
  13 Naum 4.6 x64                     :  2869.20    778    57.7  189   520    69   449.0   66.8  13.88  2813.53  14.23   53.1
  14 Chiron 2.0 x64                   :  2855.28    672    54.8  145   447    80   368.5   66.5  15.29  2820.14  14.33   53.3
  15 iCE 3.0 v658 POP x64             :  2855.07    647    54.0  131   437    79   349.5   67.5  14.74  2825.77  14.37   51.8
  16 SmarThink 1.80 AVX x64           :  2852.97    820    56.3  176   571    73   461.5   69.6  13.25  2808.47  14.18   52.8
  17 Texel 1.05 x64                   :  2852.19    778    53.6  166   502   110   417.0   64.5  13.47  2826.51  14.45   52.5
  18 Senpai 1.0 SSE42 x64             :  2835.21    740    53.1  154   478   108   393.0   64.6  14.00  2813.08  14.25   53.3
  19 Hiarcs 14 WCSC w32               :  2825.81    727    51.8  117   519    91   376.5   71.4  14.18  2813.15  14.12   52.5
  20 Shredder 12 x64                  :  2825.10    735    51.8  125   511    99   380.5   69.5  14.34  2812.87  14.23   53.6
  ---------------------------------------------------------------------------------------------------------------------------
  21 DiscoCheck 5.2.1 x64             :  2818.07    981    51.6  163   686   132   506.0   69.9  11.94  2807.10  14.20   53.1
  22 Spike 1.4 Leiden w32             :  2815.75    734    50.9  135   477   122   373.5   65.0  14.74  2809.88  14.06   52.3
  23 Booot 5.2.0 x64                  :  2815.53    805    50.4  151   510   144   406.0   63.4  13.20  2812.24  14.20   53.2
  24 Junior 13.3.00 x64               :  2814.45    687    50.6  134   427   126   347.5   62.2  14.40  2810.63  14.15   52.5
  25 Fizbo 1.6 x64                    :  2811.99    697    49.5   96   498   103   345.0   71.4  14.63  2815.42  14.26   52.7
  26 Vajolet2 2.1 POP x64             :  2808.99    823    49.3  130   552   141   406.0   67.1  13.16  2814.31  14.26   53.3
  27 Scorpio 2.77 JA POP x64          :  2802.97    653    50.6   93   475    85   330.5   72.7  14.66  2799.20  14.07   50.3
  28 Bobcat 7.1 x64                   :  2802.62    698    49.6  117   459   122   346.5   65.8  14.18  2805.88  14.10   52.2
  29 Tornado 5.0 SSE4 x64             :  2801.93    690    49.6  108   468   114   342.0   67.8  13.78  2806.16  14.07   51.9
  30 Zappa Mexico II x64              :  2793.29    805    47.7   99   570   136   384.0   70.8  13.10  2809.96  14.20   53.0
  31 Cheng 4.39 x64                   :  2789.75    773    46.2  102   511   160   357.5   66.1  13.29  2816.79  14.23   53.0
  32 EXchess 7.88b x64                :  2787.66    716    47.3  111   456   149   339.0   63.7  13.78  2807.45  14.13   52.5
  33 Doch 1.3.4 JA x64                :  2785.57    690    46.2   99   439   152   318.5   63.6  14.45  2814.08  14.21   53.5
  34 Arasan 18.3 POP x64              :  2783.32    825    46.0  109   541   175   379.5   65.6  13.06  2812.89  14.25   53.1
  35 Quazar 0.4 x64                   :  2782.77   1025    45.9  101   739   185   470.5   72.1  11.93  2812.83  14.26   53.6
  36 Dirty 03NOV2015 POP x64          :  2782.29    822    46.0  115   526   181   378.0   64.0  12.96  2811.73  14.24   53.0
  37 Sjeng c't 2010 w32               :  2781.29    699    45.7   98   443   158   319.5   63.4  14.52  2812.22  14.13   52.1
  38 Crafty 25.0.1 JA POP x64         :  2780.04    766    46.7   90   536   140   358.0   70.0  13.55  2804.21  14.17   52.1
  39 Hakkapeliitta 3.0 x64            :  2776.86    751    44.3   88   489   174   332.5   65.1  13.94  2818.41  14.24   53.0
  40 Deuterium 14.3.34.130 POP x64    :  2774.17    647    44.4   82   410   155   287.0   63.4  14.76  2814.92  14.27   53.2
  ---------------------------------------------------------------------------------------------------------------------------
  41 Pedone 1.2 BMI2 x64              :  2774.02    865    45.8  107   579   179   396.5   66.9  12.78  2804.58  14.16   51.5
  42 The Baron 3.29 x64               :  2772.91    592    46.3   71   406   115   274.0   68.6  15.53  2800.89  14.11   50.5
  43 Gaviota 1.0 AVX x64              :  2771.37    724    45.2   93   468   163   327.0   64.6  13.54  2806.76  14.17   52.1
  44 Atlas 3.80 x64                   :  2769.50    745    44.4   89   484   172   331.0   65.0  14.15  2810.60  14.12   52.5
  45 GNU Chess5 5.60 x64              :  2767.65    702    44.7   77   473   152   313.5   67.4  14.35  2806.87  14.17   52.4
  46 TogaII 280513 Intel w32          :  2766.22    734    44.2   84   481   169   324.5   65.5  13.81  2808.64  14.14   52.8
  47 DisasterArea 1.54 x64            :  2765.10    766    44.5   81   520   165   341.0   67.9  13.07  2805.34  14.14   51.7
  48 Spark 1.0 x64                    :  2762.39    777    43.6   75   527   175   338.5   67.8  14.53  2809.40  14.12   52.6
  49 Thinker 5.4d Inert x64           :  2761.60    754    43.2   72   507   175   325.5   67.2  13.98  2811.52  14.21   52.3
  50 Rodent II 0.8.7 x64              :  2756.80    682    43.1   79   430   173   294.0   63.0  14.52  2807.36  14.13   52.3
  51 Glaurung 2.2 JA x64              :  2746.43    711    41.2   53   480   178   293.0   67.5  14.11  2811.13  14.14   51.9
  52 Murka 3 x64                      :  2743.02    740    40.9   77   452   211   303.0   61.1  14.19  2809.39  14.11   52.3
  53 Nemo 1.01 Beta POP x64           :  2742.72    719    40.9   66   456   197   294.0   63.4  14.59  2809.18  14.11   52.4
  54 BugChess2 1.9 POP x64            :  2740.17    640    41.3   66   397   177   264.5   62.0  15.17  2804.19  14.07   50.8
  55 Octochess r5190 SSE4 x64         :  2733.52    698    40.3   60   443   195   281.5   63.5  15.18  2804.79  14.10   51.7
  56 Rhetoric 1.4.3 POP x64           :  2732.36    707    39.8   64   435   208   281.5   61.5  14.83  2806.99  14.08   51.6
  ---------------------------------------------------------------------------------------------------------------------------

  White advantage = 9.15 +/- 1.41
  Draw rate (equal opponents) = 70.62 % +/- 0.35
  
  Games        : 20470 (finished)
  White Wins   : 3582 (17.5 %)
  Black Wins   : 3366 (16.4 %)
  Draws        : 13522 (66.1 %)
  Unfinished   : 0

  White Score  : 50.5 %
  Black Score  : 49.5 %

[Rebel]

Hi Frank,

Hi Ed,
Interesting is:
In endgames the Elo range between the TOP-55 is ...
ONLY 244 Elo !!!!!

I believe it and much depends on the start positions. Look at the ELO of the sub-pool.

Code: Select all

   # PLAYER            :  RATING  POINTS  PLAYED   (%)    ELO
   1 WaDuuttie         :  2951.7    75.0     126    60    -5
   2 Texel 1.06a45     :  2942.1    73.0     126    58    -73 
   3 Hannibal1.5x64    :  2929.1    70.0     126    56    -117
   4 Zurichess         :  2755.2    34.0     126    27    +194

Also with my choices for the 21 start positions it simply favors weaker rated engines

[Frank Quisinsky]

Hi Ed,

yes, much weaker engines can make the points vs. Elo-Gigantics and it's nice to see that.

Shortly:
I think my endgame list is fine but not perfect !!!
My Rating List is with 4-pieces endgame table-bases ... the problem!

Best solution is to start a Rating-List without resign mode (have it) and without endgame databases (the mistake I made). But it's to late for it to changed that.

So, what I produced with move-number calculation is a good indication for "good or better" ... in endgames, not more not less.

Have fun with CENT.
The idea is good.

Best
Frank

[Maarten Claessens]

Winner of the Sub-Pool is WaDuuttie, congrats Maarten.

Thank you Ed for the tournament. I'll try to find some time to interpret the results.

[Rebel]

CENT - MAIN Pool started

Start ELO ratings according to the CCRL 40/40 rating list (1 cpu) of April 29, 2016.

Code: Select all

Stockfish     3250
Komodo        3250
Gull          3126
Equinox       3095
Andscacs      3095
Rybka         3088
Nirvanachess  3069
iCE           3050

[Rebel]

Reviewing some of the positions.

[d]8/5pk1/r5pp/P7/3R3P/6P1/5PK1/8 w - - 0 1
1927 Alekhine won the 34th game in the battle vs Capablanca to become World Champion. He did it by playing 1.Ra4! in the above position. Black's rook has to block the pawn and white has at his disposal an endless reserve of rook moves along the a-file. Again, the future for black spells "Zugzwang". The general rule is that a passed pawn should be supported by a rook from the rear. That's why ideas of 1.Rd5? are wrong. This is a knowledge that have to be implemented in many chessprograms. I have seen numerous examples of trying to support such a pawn sideways. What would your favourite program play? -- Annotator Sune Larsson

------

None of the engines played 1.Ra4 but 1.Rd5 instead and yet the 12 games ended in favor for with with a 10½-1½ score. This begs the question if the game is a technical draw after 1.Rd5

[Maarten Claessens]

SF-7 with tablebases:

Code: Select all

 30/47	04:38	 152,359,672	548,324	+1.84	Rd4d5 Ra6c6 Kg2f3 Kg7f6 h4h5 Kf6e7 Rd5b5 Rc6a6 Kf3e4 Ke7d8 Ke4d4 Kd8d7 Rb5b7+ Kd7e8 h5xg6 f7xg6 Rb7b5 Ke8d7 f2f4 Ra6a8 Kd4c4 Ra8a6 Kc4b4 Kd7c7 Rb5e5 Ra6d6 Kb4b5 h6h5 a5a6 Kc7b8 Kb5a5 Rd6d3 Re5g5 Rd3a3+ Ka5b6

[Rebel]

[d]8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 1
None of the engines did find Shirov's elegant 1...Bh3 giving withe at least some hope for a draw. Curious what the big-boys will do here.