Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 3000 ELO ?

Author: Dann Corbit

Date: 13:02:56 07/31/01

Go up one level in this thread


On July 31, 2001 at 15:46:27, William Wentworth wrote:
>Can the level of 3000 or more ELO be reached on multiple processor systems that
>currently exist ?
>Or should we wait till the new crystal-laser processors get commercialized ?
>When will that hardware be available ? Only Japanese ?
>Any estimates about its price ?

3000 ELO is completely meaningless unless you specify precisely the competition.
Against all the kindergarten children in the US, the weakest chess engines have
an ELO of 3000.

If you could get a program that was 3000 ELO FIDE, it would be no different from
one that was 2900 ELO FIDE.  A human (besides perhaps Kasparov, Kramnik, Anand)
will have no idea which one he/she is playing because they will lose all the
games.

The number is pretty well meaningless to most people anyway.  The only thing
that matters in an ELO rating is the difference between that number and other
numbers in the pool.

If you have a pool of talent that are rated at 2000 ELO on average, and then you
suddenly just give every member 500 ELO, the predictions of the ELO system will
not change at all.

It is completely driven by distance between opponents.  The absolute numbers are
actually quite meaningless (e.g. 3000 might be weaker than 2500 if you compare
different pools of talent).

Look at this, where ELO ratings were calculated with an initial assumption of
2300 ELO average:

    Program          Score     %    Av.Op.  Elo    +   -    Draws

  1 TSCP         :  30.0/ 36  83.3   2275   2555   83 155   22.2 %
  2 Monik        :  25.0/ 36  69.4   2289   2432   96 146   11.1 %
  3 Storm        :  21.5/ 36  59.7   2297   2365  107 112   19.4 %
  4 Noonian      :  21.0/ 36  58.3   2298   2356  109  78   44.4 %
  5 Ozwald       :  19.5/ 36  54.2   2300   2330  115 104   19.4 %
  6 Zephyr       :  19.5/ 36  54.2   2300   2330  115  97   25.0 %
  7 LarsenVB     :  19.5/ 36  54.2   2300   2330  115  97   25.0 %
  8 SnailSCP     :  10.0/ 36  27.8   2320   2154  129  93   22.2 %
  9 Raffaela     :   9.0/ 36  25.0   2322   2132  135  90   22.2 %
 10 Golem01      :   5.0/ 36  13.9   2335   2018  237  81    5.6 %


    Program          Score     %    Av.Op.  Elo    +   -    Draws

  1 Freyr        :  31.0/ 44  70.5   2292   2443   84 130   13.6 %
  2 Cilian       :  30.5/ 44  69.3   2293   2434   85 131   11.4 %
  3 Sjeng73      :  30.0/ 44  68.2   2293   2426   86 117   18.2 %
  4 Faile        :  28.0/ 44  63.6   2296   2394   91 102   22.7 %
  5 Averno       :  25.0/ 44  56.8   2301   2348   99  93   22.7 %
  6 SSEchess     :  24.0/ 44  54.5   2302   2334  103  90   22.7 %
  7 Gully2       :  22.5/ 44  51.1   2304   2312  108  94   15.9 %
  8 ColChess     :  22.0/ 44  50.0   2305   2305   95  95   22.7 %
  9 Chessterfield:  16.5/ 44  37.5   2312   2223   97  92   25.0 %
 10 DChess       :  16.5/ 44  37.5   2312   2223  123  92    6.8 %
 11 NewRival     :  12.0/ 44  27.3   2319   2148  126  82   18.2 %
 12 Crux         :   6.0/ 44  13.6   2331   2011  204  71    9.1 %


    Program          Score     %    Av.Op.  Elo    +   -    Draws

  1 ExChess      :  32.0/ 44  72.7   2285   2455   82 126   18.2 %
  2 ArasanX      :  31.0/ 44  70.5   2287   2438   84 106   27.3 %
  3 lambChop     :  29.5/ 44  67.0   2289   2412   87 104   25.0 %
  4 Ant          :  25.5/ 44  58.0   2295   2350   98  97   20.5 %
  5 InmiChess3   :  24.0/ 44  54.5   2297   2328  103  86   27.3 %
  6 Knightx      :  24.0/ 44  54.5   2297   2328  103 106    9.1 %
  7 GNUChess4    :  19.5/ 44  44.3   2303   2263   89 101   25.0 %
  8 Dragon       :  18.5/ 44  42.0   2304   2248   97  98   20.5 %
  9 Amyan        :  16.0/ 44  36.4   2307   2210   90  91   31.8 %
 10 gnuchess     :  16.0/ 44  36.4   2307   2210  102  91   22.7 %
 11 Fortress     :  14.0/ 44  31.8   2310   2178  117  86   18.2 %
 12 LDBlanche    :  14.0/ 44  31.8   2310   2178   95  86   31.8 %


    Program          Score     %    Av.Op.  Elo    +   -    Draws

  1 LGoliath     :  37.0/ 52  71.2   2289   2445   76  99   26.9 %
  2 Crafty       :  32.0/ 52  61.5   2294   2376   85  77   34.6 %
  3 AnMon        :  31.5/ 52  60.6   2294   2369   86  83   28.8 %
  4 Amy          :  28.5/ 52  54.8   2297   2331   93  85   21.2 %
  5 SOS          :  28.0/ 52  53.8   2298   2325   94  82   23.1 %
  6 Bringer      :  27.5/ 52  52.9   2298   2318   96  72   32.7 %
  7 TCBishop     :  27.0/ 52  51.9   2299   2312   97  73   30.8 %
  8 Phalanx      :  26.5/ 52  51.0   2299   2306   99  74   28.8 %
  9 Comet        :  24.5/ 52  47.1   2301   2281   62  96   44.2 %
 10 Gromit3      :  24.5/ 52  47.1   2301   2281   79  96   25.0 %
 11 Gromit2      :  21.5/ 52  41.3   2304   2243   89  88   21.2 %
 12 Francesca    :  18.5/ 52  35.6   2307   2204  102  82   17.3 %
 13 Yace         :  18.5/ 52  35.6   2307   2204   87  82   28.8 %
 14 ZChess       :  18.5/ 52  35.6   2307   2204  108  82   13.5 %


Notice that TSCP is stronger than Little Goliath, Crafty, AnMon, Amy, SOS, etc.
Of course, we all know that is absurd.  TSCP is a tiny test program of 50K and
does not even have an opening book, piece list, or a hash table.  But it played
in a different pool of talent.  Since the pools are segregated, you cannot make
any inference based upon the numbers in the different regions.  Only within the
region for which it was calculated.




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.