Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fruit 2.1 with and without Delta & Futility pruning compared

Author: Heinz van Kempen

Date: 04:58:29 06/24/05

Go up one level in this thread


On June 24, 2005 at 03:54:23, M Hurd wrote:

>On June 23, 2005 at 22:59:14, Aaron Gordon wrote:
>
>>Ran two small tournaments to see if I could discern any difference between the
>>stock Fruit and one with Delta and Futility pruning enabled. I wish I had more
>>time, this is all I could do for now. I'll let you be the judge of the results
>>and games.
>>
>>Settings:
>>Pentium 3 1.2GHz, ponder off, 64mb hash per engine, time control was 3 1,
>>Perfect v6.1 book was used and set to "optimize".
>>
>>Fruit v2.1 with no changes:
>>
>>34.0 Shredder 9
>>30.5 Fritz 8 Bilbao
>>30.0 Fruit 2.1
>>25.5 Junior 9
>>
>>Download the games from that tournament here:
>>ftp://spdycpu.dyndns.org/f21default-tourney.zip
>>
>>
>>
>>Fruit 2.1 with Delta and Futility pruning enabled:
>>
>>33.5 Fruit 2.1
>>33.5 Shredder 9
>>32.0 Fritz 8 Bilbao
>>21.0 Junior 9
>>
>>Download the games from that tournament here:
>>ftp://spdycpu.dyndns.org/f21delta-tourney.zip
>>
>>Not definitive proof of anything, but I think it is starting to show a small
>>trend that delta and futility pruning helps.
>
>
>From the few games I have played it does indeed look as though delta and
>futility pruning may help. It would be great if the CEGT could test this, we
>could then compare the results with the standard version which they have played
>around 800 games so far.
>
>Regards
>
>Mike

Hello,

thanks for information.

CEGT testers will of course first give 2000 games with default settings to have
decent statistical values. After this we will wait for more results without
Delta and Futility pruning and then contact Fabien if he thinks that another
test would be useful. And finally like always in CEGT there will be a vote from
all testers and the majority will decide if they want to run another big test,
as the same amount of games with same conditions against the same opponents
would be needed. To be very precise even on the same machines.

Maybe we will vote doing this, maybe not, because there are also a lot of other
engines that need more tests.

Best Regards
Heinz



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.