Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Congratulation for chesstiger(probably right but too early)

Author: Uri Blass

Date: 10:11:33 08/24/01

Go up one level in this thread


On August 24, 2001 at 12:00:14, José de Jesús García Ruvalcaba wrote:

>On August 24, 2001 at 10:50:36, Uri Blass wrote:
>
>>On August 24, 2001 at 10:34:01, José de Jesús García Ruvalcaba wrote:
>>
>[big snip]
>>>
>>>Hi Uri,
>>>plese try the following experiment with elostat.
>>>1. Players A, B, and C play each other, with the following individual results:
>>>A beats B 99.5 to 0.5
>>>B beats C 99.5 to 0.5
>>>A beats C 100 to 0
>>>Which ratings do you get for A, B and C using Elostat?
>>>
>>>2. The same players, but with the following results:
>>>A beats B 99.5 to 0.5
>>>B beats C 99.5 to 0.5
>>>Same question as for part 1.
>>>
>>>If the program behaves correctly, the rating of A for part 1 should not be lower
>>>as the rating of A for part 2.
>>>José.
>>
>>Unfortunately the program needs pgn and it calculate the results unless it is a
>>competition by 2 players.
>>
>>Here is some information from the readme file of this program
>>
>>Following this theory, the Elo rating corresponding to a relative performance of
>>100 % or 0 % is indefinite. Due to mathematical reasons (e.g. to guarantee the
>>feasibility of the iteration procedure) ELOStat assigns to those programs a
>>finite Elo value which is exactly 600 points smaller (0 % perf.) or greater (100
>>% perf.) than the Av.Op. Elo. Or in other words: ELOStat does not support Elo
>>differences greater than  600 points (therefore the 95% error margins
>>can be at most  1200 points). For nearly all practical purposes, this
>>restriction does not play an important role.
>>
>>In very rare cases ELOStat produces an error message stating that the iteration
>>procedure failed and that no convergence of the Elo mean value could have been
>>reached within the maximum number of iterations specified by the program. This
>>problem only appears when many programs in the database are characterized by 0 %
>>or 100 % results. In these cases the iteration procedure is slowed down
>>significantly so that the Elo calculation takes a much longer time as usual.
>>
>>Uri
>
>Ok, if it needs a pgn file, then please try the following experiment.
>Remove the game between Junior and Ruy López from your WMCCC pgn file, and use
>Elostat with this new file. Now compare Junior's rating from the complete and
>the incomplete file. If the program behaves correctly, the rating for Junior
>with the complete file should not be lower than the rating with the incomplete
>file (and Ruy López' should not be higher).

You are right
see my other post for the results without that Junior game.

Uri



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.