Gromit 2.20

[Andrew Tanner]

Geschrieben von:/Posted by: Andrew Tanner am 16. Dezember 1999 21:06:49:

For all the winboard engine testers:
I have found that if you lower pawnhashsize well below it's minimum of 8 to a value of lets say 3.. it plays a very strong tacticall game in Blitz. Maybe if someone would like to test this at longer time controls we will see if there is any merit to this
Have Fun,
A. Tanner

[Imran Hendley]

Geschrieben von:/Posted by: Imran Hendley am 16. Dezember 1999 21:39:50:
Als Antwort auf:/As an answer to: Gromit 2.20 geschrieben von:/posted by: Andrew Tanner am 16. Dezember 1999 21:06:49:

For all the winboard engine testers:
I have found that if you lower pawnhashsize well below it's minimum of 8 to a value of lets say 3.. it plays a very strong tacticall game in Blitz. Maybe if someone would like to test this at longer time controls we will see if there is any merit to this
Have Fun,
A. Tanner

Very interesting. I have observed that Gromit, while very good at long games, is sometime horrible at Blitz. I will try lowering the pawnhashsize and see what happens. Also, if anyone reading this has a special eval file for Gromit that they use in Blitz, or that they think improves its Blitz strenght, please post it or email it to me. Thanks.

[Frank Schneider]

Geschrieben von:/Posted by: Frank Schneider am 18. Dezember 1999 08:51:14:
Als Antwort auf:/As an answer to: Gromit 2.20 geschrieben von:/posted by: Andrew Tanner am 16. Dezember 1999 21:06:49:

For all the winboard engine testers:
I have found that if you lower pawnhashsize well below it's minimum of 8 to a value of lets say 3.. it plays a very strong tacticall game in Blitz. Maybe if
someone would like to test this at longer time controls we will see if there is any merit to this
Have Fun,
A. Tanner

There are three possible reasons for this:
a) clearing the hashtables takes too much time (very unlikely)
b) a serious bug (size of pawnhashtable should only influece search*speed* and only very slightly
c) you haven't played enough games and the effect you observed isn't there
b) is unlikely because it would also influence games at slower timecontrols.
How many games did you play?
Frank