Computer Chess Club Archives


Search

Terms

Messages

Subject: Battle of the Crowns -- calibration data [are we there yet?]

Author: Dann Corbit

Date: 11:58:49 06/27/00


Program         Elo    +   -   Games   Score   Av.Op.  Draws
-------------  ------ --- ---  ----    ------   ----   ------
Crafty        : 2516   16  22  1073    67.3 %   2391   25.4 %
SOS           : 2441   25  29   533    60.3 %   2368   21.6 %
Comet         : 2431   17  17  1257    56.6 %   2385   23.9 %
LGoliath      : 2430   23  19   822    53.4 %   2406   26.5 %
AnMon         : 2411   24  20   803    52.4 %   2394   23.4 %
Patzer        : 2397   23  29   564    49.7 %   2398   23.6 %
Phalanx       : 2396   19  22   943    49.0 %   2403   19.9 %
Gromit3       : 2381   41  34   285    50.7 %   2377   21.4 %
TCBishop      : 2378   21  24   768    47.9 %   2392   21.6 %
Amy           : 2359   25  36   486    64.6 %   2254   13.6 %
Gromit2       : 2358   25  27   545    44.1 %   2399   24.4 %
Francesca     : 2340   24  27   572    45.2 %   2374   25.3 %
ZChess        : 2331   21  25   701    47.6 %   2348   23.1 %
Yace          : 2327   35  30   367    52.7 %   2308   23.7 %
Bringer       : 2311   26  30   499    48.5 %   2321   21.2 %

Program         Elo    +   -   Games   Score   Av.Op.  Draws
-------------  ------ --- ---  ----    ------   ----   ------
ArasanX       : 2287   37  31   344    51.5 %   2277   20.3 %
LambChop      : 2279   41  34   291    38.3 %   2362   19.6 %
Ant           : 2256   31  30   436    54.5 %   2225   18.6 %
GnuChess4     : 2256   65  67    99    46.5 %   2280   14.1 %
InmiChess3    : 2245   50  45   187    51.9 %   2232   16.0 %
ExChess3      : 2226   26  32   507    60.4 %   2153   14.2 %
Knightx       : 2216   49  57   146    48.6 %   2225   19.2 %
InmiChess2    : 2199   96  76    59    34.7 %   2308   18.6 %
Dragon        : 2166   59  36   212    30.0 %   2314   15.6 %
GnuChess5     : 2155   50  66   130    64.2 %   2053   16.2 %
Fortress      : 2144   67  68    92    56.0 %   2103   16.3 %
LDBlanche     : 2131   57  42   170    34.1 %   2246   20.0 %
Amyan         : 2115   57  58   135    47.8 %   2131    9.6 %

Program         Elo    +   -   Games   Score   Av.Op.  Draws
-------------  ------ --- ---  ----    ------   ----   ------
Gully2        : 2062   59  64   119    49.6 %   2065   10.1 %
Cilian        : 2051   46  40   213    39.0 %   2129   20.7 %
ColChess      : 2044   65  69    92    58.2 %   1986   18.5 %
SSEChess      : 2010   81  49   114    29.4 %   2163   14.9 %
Sjeng         : 2002   48  42   205    40.2 %   2071   17.1 %
Averno        : 1992   61  34   239    31.0 %   2131    7.5 %
Freyr         : 1946   63  97    83    64.5 %   1843    3.6 %
Crux          : 1938  101  96    46    42.4 %   1991   15.2 %
DChess        : 1920   77  57   109    39.0 %   1998    6.4 %
NewRival      : 1911   87  68    73    35.6 %   2014   16.4 %
Faile         : 1899   96  66    76    34.9 %   2007    9.2 %
Monik         : 1868   73  78    71    44.4 %   1908   21.1 %
Chessterfield : 1864   65  20   499    19.7 %   2107    4.2 %
TSCP          : 1801   52  48   172    42.7 %   1852   13.4 %

Program         Elo    +   -   Games   Score   Av.Op.  Draws
-------------  ------ --- ---  ----    ------   ----   ------
Zephyr        : 1757  168  79    42    21.4 %   1983    9.5 %
SnailSCP      : 1755  234  78    37    12.2 %   2099    8.1 %
Ozwald        : 1720   68  66    91    42.3 %   1774   18.7 %
Noonian       : 1684  158  47    90    14.4 %   1993    6.7 %
Storm         : 1589  142  56    77    22.1 %   1809    2.6 %
LarsenVB      : 1565  292 185    11     9.1 %   1965   18.2 %
Golem01       : 1397  242  64    49     8.2 %   1817    8.2 %
Raffaela      :  992    0   0     3     0.0 %   1592    0.0 %

Well, we are drawing close to the wire. Calibration of the bottom programs is
very problematic. Most of them are completely deterministic, and so playing a
set of ten games only produces two distinct games. For instance, I ran 10
Raffaela games yesterday against TSCP and none of them were different than the
ones already stored. Since I filter for duplicates, the net effect of those ten
games was zero!

In the actual contest, if you lose the same way four times in a row, you still
lost the games.

Just call me "no mercy" corbit.

It might seem like a good thing to squeak into the next higher bracket above
you.  However, if you keep in mind that the overall loser from each bracket gets
demoted and the overall winner from each bracket gets promoted, it might be
better for a "crowns" rating to stay put in a lower bracket.

On the other hand, TSCP is markedly improved with the new and simple timing
changes.  Yace and Bringer are both much stronger with recent changes and I
predict that none of these three mentioned programs will finish last in their
bracket, despite their lowly starting positions.

I think also that ArasanX will probably win its bracket as well, but the new
GnuChess might give it a run for its money (there is a completely new bitboard
version of GnuChess5).







This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.