Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Time controls for engine v engine matches.

Author: Bertil Eklund

Date: 15:43:38 05/04/01

Go up one level in this thread


On May 04, 2001 at 13:47:45, Uri Blass wrote:

>On May 04, 2001 at 13:26:35, Chris King wrote:
>
>>Although there is a huge difference between blitz chess and normal chess for
>>humans isn't it true that for computer programs, that don't fell pressure, the
>>only difference is the depth of search. After all programs are analysing
>>millions of positions per move even at blitz chess on today's computers. So
>>engine v engine results played at blitz should only reflect results played at
>>longer time controls. I have often seen it stated that certain engines play
>>better at longer time controls, but is this really so. I have also seen it
>>stated that at longer time controls an engine has more time to formulate a plan.
>>If this were true the nodes per second count would significantly reduce as an
>>engine looked deeper into a position and made it's more sophisticated plan. I
>>have never noticed such a reduction in nodes per second and isn't it true that a
>>computer program doesn't make plans in any human sense.
>
>I read that tiger makes plan but
>I guess that making the plan takes a very small time so you cannot see reduction
>in nodes per second because of the plan.
>
>I guess that it is using the root position to decide about some targets and
>changes the evaluation function based on these targets.
>
>I have no reason to believe that this help tiger more at longer time control and
>the data that I read suggests that Deep Fritz earns more than Tiger from
>time(Tiger is clearly better at blitz when it is not clear if tiger is better
>than Deep Fritz at tournament time control)
>
>>You can play far more engine v engine games at blitz which gives a statistically
>>more accurate measurement of the relative strength of each engine, or am I
>>missing something glaringly obvious here.
>
>Yes
>You are missing something.
>
>Programs have different evaluation function and different search rules.
>The best evaluation function for blitz is not the best evaluation function for
>long time control.
>The best search rules for blitz are not the best search rules for long time
>control.
>
>Uri

Hi!

Yes, you are right.

I believe that the Tigers are superior in blitz/speed-chess but at 2h/40 I am
still not absolutely sure. My guess is that Tiger could be slightly better then
Deep-Fritz at 2h/40 but I'm not sure. In the next list I hope we can include at
least Tiger(s) and Deep Fritz on Athlon 1200mhz.

Bertil



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.