Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Between 2 versions

Author: Dave Kuntzsch

Date: 18:52:51 08/04/02

Go up one level in this thread


On August 04, 2002 at 21:35:06, Dave Kuntzsch wrote:

>On August 04, 2002 at 21:19:00, TEERAPONG TOVIRAT wrote:
>
>>
>>Hi,
>>
>>1. Suppose the newer version of program A beats older version comfortably.
>>But their results with program B are approximately equal.
>>All are statistically significant. Can I conclude that the newer version
>>is slightly better? Or I have to test with program C,D,E...?
>>
>>2. From testing positions, the newer version spends more time than previous
>>one about 10% but it produces less searching nodes 10%. Both get  the same
>>score. I think it is possible that the newer version may be better in the long
>>run especially total consuming time. Is it true?
>>
>>Thanks,
>>Teerapong

Sorry about the null post above. My mouse finger twitched. I have been running a
five minute blitz tournament for many months involving various versions of
Crafty, Anmon 5.21, and Yace 0.99.56. Each round of four engines involves 150
games against each of the opponents. There are eight rounds in all with the same
four opponents for a grand toal of 1200 games played by each engine. I keep
Anmon and Yace as constants and rotate in two Craftys in each round of eight.
The Craftys are 17.14, 18.14, both versions of Dann Corbit's 18.15 and Robert
Hyatt's 18.15. Right now I'm halfway through the last iteration. Then I'll keep
the best performing Crafty and and replace the other with Comet B.48. Then start
a new tournament. BTW, Anmon is doing very well, and so far Yace is the cream of
the crop.

As for your test, I would have both engines play multiple opponents many games
before I could say one is better than the other.

Dave



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.