Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 3 computer chess myths: which one has proven to be true?

Author: Uri Blass

Date: 06:50:07 04/22/02

Go up one level in this thread


On April 22, 2002 at 09:24:12, David Dory wrote:

>On April 22, 2002 at 09:01:30, Jouni Uski wrote:
>
>>1. program X plays relatively better against humans than against computers
>>2. program Y plays relatively better with longer time control
>>3. there are diminishing returns from speed doubling, when search depth is
>>   increased
>>
>>I think definitely 3. is true - only the size of diminishing is unclear. E.g.
>>from latest SSDF list (yes comp-comp play!) doubling gives only 40 points at
>>current top level against 75 previously. But 1 and 2 are still open cases.
>>
>>Jouni
>
>#1 in addition to a more hand-tuned evalu8, better play against humans requires
>a better book. Not for an average player, but to take on a GM who knows the
>latest busts in ECO, you need a very up to date book or the program is behind
>right from the start.
>
>I think Rebel has been a great example of this.
>
>#2 True. You can program a version of your engine just for blitz and make it
>more effective. Hiarcs would NOT be a blitz champ - it isn't designed for it.
>The 17XX Crafty and Fritz6 are more designed for blitz than either Hiarcs or the
>18XX Crafty, IMO. They just do more nps, and that's not the only thing in blitz,
>but it's a bigger part of the game than in normal time control.


I disagree.
nps mean nothing.

Hiarcs7.32 is probably worse at long time control thanks to worse branching
factor.

I guess that things are going to be different with Hiarcs8.

Fritz7 is also known to be better than Fritz6 in all time controls and the
fact that Fritz6 counts more nps does not change it.

Uri



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.