Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: news from gandalf......

Author: Jason Williamson

Date: 19:50:44 09/22/00

Go up one level in this thread


On September 22, 2000 at 16:48:08, Chessfun wrote:

>On September 22, 2000 at 14:38:07, Jason Williamson wrote:
>
>>Not to much differnt from the default, I just add nul_low 4, and nval 1.  I also
>>do:
>>
>>resign
>>with the threshold -9 for 5 moves.
>>
>>My opinion on null_low 4 is that its better at faster time controls, weaker at
>>slower.  But its still seems to be doing ok on the ICC regardless.  Wonder what
>>would happen if i tried null_low 5 ;).
>
>I ran null_low 5 for a while and again found that it was weaker than the
>default.
>
>>I use nval 1 because Yace was pitching 2 minors for a r and p way to much.  I
>>don't see that as much anymore and as a result, Yace plays much stronger.
>
>Could be null_low 4 combined with nval 1 is stronger?
>
>>I built a book out of 510,000 games.  Its only 110 megs in size ;).
>
>Biggest I could manage was 70 Mb's. I did at one point get a 112 Mb
>book out of a 1 GB database but could not run more than 20 Ply due
>to processor and ram so I stayed with the 70 Mb book.
>
>>I did a test run with null_low 4 vs 10 other engines, the resulting rating for
>>null_low 4 landed up being 40 points higher then the default.  Of course, I
>>figure it needs more testing to be sure, but as I said, I suspect that null_low
>>4 is good for quick time controls.
>
>Again the Yace at my Young Talents + blitz was with null_low 4 and when I ran
>separate against the others it did much better. That was pre 5 BTW.
>
>Sarah.
>
>>JW

I was using pre 18+.



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.