Author: Ulrich Tuerke
Date: 09:24:36 01/17/02
Go up one level in this thread
On January 17, 2002 at 11:29:42, Uri Blass wrote: >On January 17, 2002 at 11:00:44, Ulrich Tuerke wrote: > >>On January 17, 2002 at 01:30:31, Harald Faber wrote: >> >>>After having changed the hash size from 200MB to 56MB due to some >>>recommendations here (104MB hash are recommended for Athlon-1000, so 56MB should >>>be the right size for my Athlon-500), Gandalf 5 performed much better in this >>>match and reached a 4.5-5.5. Although the start was all but convincing with >>>1.0-4.0. One does not forget here that Hiarcs 7.32 is a bit weaker and less >>>dangerous opponent than Junior 7, Shredder 6 and Fritz 7a, it is still unclear >>>whether Gandalf plays better with this hash size. IMO this should not make such >>>a difference, but who knows? Electricians say "Women and currency go strange >>>ways." :-) >>> >>>Next match starting in a few minutes is vs. Fritz 6a. >> >>I think, you shouldn't assume a causal relation between shrinking the hash table >>size and the results after this. >> >>Assume you would have gone out to wash your hands, and after you're back Gandalf >>does the turnover. Would you think that washing your hands was the reason for >>the improvement. -:) >> >>We shouldn't be fooled by statistical fluctuations. >> >>Uli > >It is different. > >It is clear that washing your head does not change the strength of the program. >It is not clear that gandalf has no bug with too big hash tables. > >We saw no proof that gandalf is better with lower hash tables but we saw enough >to suspect that it may be the reason. > >Uri I just think the chance, that Harald had just observed a statistical fluctuation is overwhelmingly large. Of course you can't exclude with 100 percent certainty some reaL peculiar program bug. Uli
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.