Author: Christophe Theron
Date: 18:05:04 05/01/00
Go up one level in this thread
On May 01, 2000 at 10:37:06, Enrique Irazoqui wrote:
>On April 30, 2000 at 22:43:01, Christophe Theron wrote:
>
>>On April 30, 2000 at 18:48:34, Chris Taylor wrote:
>>
>>>If the trend carries on? Crafty does better at longer time controls??
>>>The more it can *fink* The better it is doing!
>>>
>>>Chris Taylor
>>
>>
>>The trend you are talking about does not exist in the presented data.
>>
>>All you can say is that Crafty has big problems at 1 and 2 mins per game. It may
>>be a problem with the interface due to the time taken by the GUI to update the
>>screen.
>>
>>Apart from that, we have to wait until the slower time controls games are
>>played.
>>
>>The match played on one computer with ponder=off showed similar variations
>>(except the obvious problem at very short time controls), but no trend at all.
>>
>>
>> Christophe
>
>I understood that the purpose of these tests was to see to what extent blitz
>results and ponder off relate to slow games with ponder on.
The purpose of the match is to see what happens when time controls varies and
pondering varies.
Sometimes blitz matches and slow time controls matches results are related, and
that tells us something about the opponents.
Sometimes they do not relate, and that tells us something else about the
opponents.
In both cases, we learn something interesting about the programs.
For example, in the case of Cratfy/Fritz, it seems that Crafty's search scales
well with Fritz' search (they are of equal quality). So the difference between
the programs is mainly in the speed in term of NPS and in the quality of the
evaluation function. That's what the match tells me so far.
> I think they don't.
I agree with you. Did I say they did?
Christophe
>A couple of nights ago I ran a tournament at 1 1, ponder off, 20 game matches on
>a P600E, 4MB hash each engine, with the following results:
>
>Hiarcs 7.32 55.5/80
>Junior 6a 44/80
>Fritz 6a 43.5/80
>Crafty 17.10 32/80
>Nimzo 7.32 25/80
>
>The results of Hiarcs and Nimzo would have been very different at 40:2 ponder
>on, Hiarcs quite a bit lower and Nimzo much higher, as shown in the SSDF list
>and in some tournaments, including my own.
>
>Enrique
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.