Author: Albert Silver
Date: 09:06:09 11/05/98
Go up one level in this thread
On November 05, 1998 at 10:00:09, Enrique Irazoqui wrote:
>On November 05, 1998 at 09:19:20, Albert Silver wrote:
>
>>I suspect the interest in this will be enormous. Can you post the evals and/or
>>depths of the programs as well?
>
>Of course, but from one of the two players only. Merging both scores would be
>too much work. Besides, in order to speed up the autoplaying by avoiding
>timeouts, I set Junior 5 and Fritz 5 to play without letting the opponents to
>save their games. Then I get one side score in many games.
>
>Small problem. In my testing I played 38 games Rebel 10 - Fritz 5. Rebel won
>20-18. Since in my tournament I play 10 game matches, should I count 5.5-4.5 for
>Rebel, which is proportional to the total result in 38 games, or the 6-4 I got
>from the first games? Half a point is no big deal, but... I think the 5.5-4.5 is
>more real.
>
>Enrique
>
>> Albert Silver
I agree. Particularly as the proportion should actually be 5-4.5, but this also
serves to show the danger of evaluating a program with a short match. Regarding
the evals, of course both scores would be best, but what is really of interest
is to see what the programs think of their own position and where they differ.
Albert Silver
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.