Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Distributed perft, current standings and trends

Author: Albert Bertilsson

Date: 23:56:26 12/12/03

Go up one level in this thread


On December 12, 2003 at 17:26:04, Aaron Gordon wrote:

>I was wondering if it may be better to track the stats by trillions of moves
>done instead of on a block by block basis. Because the current way doesn't seem
>as if it'd be completely fair. For example.. lets say computer X is twice as
>fast as computer Y. However, computer Y completes two smaller blocks in the same
>time it takes computer X to do one. Now the stats report that computer Y has
>done two blocks, looking like it is twice as fast as computer X when actually
>the reverse is true. Just something to think about. Thanks for the project, I
>always love the distributed stuff. :)

It would be hard to implement that change now, because it requires an addition
in the database, the server and a new client. If the project continues with
perft(12) I'll invest more time in all areas to make a nicer client and a better
server. I would then add top list based on MNPS instead.

Ordering by MNPS would be easier and actually favour those lucky enough to get
the large problems (large problems are searched faster because of the hash table
(specially for those with 512MB+ ht)).

I also intend to add some way to view the current position, this is a client
only upgrade.

/Regards Albert



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.