Author: Dann Corbit
Date: 19:03:09 09/02/02
Go up one level in this thread
On September 02, 2002 at 15:47:27, Dieter Buerssner wrote:
>On September 01, 2002 at 19:12:01, Andrew Williams wrote:
>
>>The final score from the point of view of the version with the book was:
>>
>> +71 =78 -51
>>
>>With a final score of Withbook 110 - 90 Nobook.
>
>This may look a bit more significant, than it really is. A Monte Carlo
>simulation of 200 game matches between equal opponents (all games independent -
>no learning) shows, that in about 10% of such 200 game matches, a result of
>110-90 or more extreme can be expected. I used a draw probability of 39% (as in
>your match), white win probability of 34% and balck winning probability of 26%.
>Some changes in these numbers will give slightly other results.
>
>A score of 220-180 and better in a 400 game match, could be expected in about 2%
>of the cases.
I have continued the experiment with crafty and have a 60% score with book
verses 40% w/o book at this point. All games are at G/90:
Program Elo + - Games Score Av.Op. Draws
1 Crafty-18.15w : 2536 81 51 58 60.3 % 2464 55.2 %
2 Crafty-18.15n : 2464 51 81 58 39.7 % 2536 55.2 %
(1) Crafty-18.15w : 58 (+ 19,= 32,- 7)
Crafty-18.15n : 58 (+ 19,= 32,- 7)
The games:
ftp://cap.connx.com/pub/chess-engines/new-approach/booktest.pgn
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.