Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Detecting three-fold repetition?

Author: Ian Osgood

Date: 13:08:03 07/19/00

Go up one level in this thread


On July 17, 2000 at 19:27:09, Dann Corbit wrote:

>On July 17, 2000 at 19:02:29, Peter Kappler wrote:
>[snip]
>>>To squeeze 2000 ELO from such a tiny package is nothing short of a miracle.
>>>There are programs with ten times as much code that get the stuffings pounded
>>>out of them.
>>
>>Is it really that strong?  My experience with it is limited, but my impressions
>>were that it was not close to that strength, at least at slow chess.  Blitz
>>might be a different story.
>>
>>(No offense intended to Tom, since the primary design goal was simplicity.)
>
>Here is the calculated ELO at Standard time control (G/60 or slower for this
>list) for the 1-Crown category:
>
>Program         Elo    +   -   Games   Score   Av.Op.  Draws
>--------------  ----   --  --  -----   ------  ------  ------
>Monik         : 1833   71  60    96    51.0 %   1826   18.8 %
>TSCP          : 1811   49  47   187    44.7 %   1848   13.4 %
>LarsenVB      : 1758  200 169    15    30.0 %   1905   20.0 %
>Zephyr        : 1743  133  67    58    23.3 %   1950   12.1 %
>SnailSCP      : 1735  111  70    60    29.2 %   1889   15.0 %
>Noonian       : 1724  125  47    99    18.7 %   1980    9.1 %
>Ozwald        : 1721   65  64    97    42.3 %   1775   20.6 %
>Storm         : 1655  120  47   109    22.5 %   1870    2.8 %
>MFChess       : 1515  502  80    21     4.8 %   2036    0.0 %
>Golem01       : 1380  196  48    79     7.6 %   1814   10.1 %
>Raffaela      : 1149    0   0     9     0.0 %   1749    0.0 %
>
>It looks from here loke TSCP is only about 1800 ELO.  But at least half of the
>187 games were made with older iterations of TSCP (1.4x) which had an ELO of
>about 1650 because it only used fixed ply for moving.  Hence, I am guessing that
>the real ELO will be 2000 or so, once we have an accurate test set.
>
>So yes, it really is fairly strong for such a tiny program.  If you compare the
>size of the binaries it is almost funny.  I am quite sure that TSCP gets the
>most ELO per byte of binary of any program on earth.

I don't know... there were some pretty tiny programs running on early
DOS/UNIX/Apple hardware.  To be fair you should include them on recent hardware.
 And then there are Richard Lang's programs...  (Genius 2 was only 85K including
the graphical DOS UI!)

Another free program worth including in your tests would be SCP (Stanback's
Chess Program), the ancestor of GnuChess.  It also compiles very small.  In
fact, it is the basis of the recent PocketChess line of PDA chess programs.  Is
SnailSCP above derived from SCP?

Ian



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.