Author: José Carlos
Date: 06:12:17 12/20/01
Go up one level in this thread
On December 20, 2001 at 07:31:02, Severi Salminen wrote:
>>96-99% is incredibly high - it's hard to believe, actually.
>
>I'd gladly post the sources, but...nope ;) I had thought that 95% is a normal
>figure but it seems to be high, I'm not complaining though! I checked the code
>and it seems to OK: I increase moves_searched after I have really searched a
>legal move (futile and illegal moves don't count as they are not searched) and
>then I increase FailHighCount and FirstMoveFail counters as necessary. And now I
>removed the code from qsearch, which might have biased the figures (well, it
>didn't). I'm using PVS and SEE (the latter of course has a great impact on move
>ordering).
>
>>Could you post the test position you're using?
>
>It's the position after 1.e4 e5 2.d4 d5:
>
>[D] rnbqkbnr/ppp2ppp/8/3pp3/3PP3/8/PPP2PPP/RNBQKBNR w KQkq - 0 1
>
>>It would also be nice to know if
>>these numbers are repeatable over a large set of positions.
>
>I have not yet implemented test suite processing, but I can test individual
>positions of course. From initial position, the percentage was 94%.
>
>Severi
Averno 0.35 from initial position (after about 30 secs in a PIII-900):
No SEE SEE
R=0 87% 87%
R=2 91% 91%
R=3/2 91% 92%
R=4/2 91% 91%
I must add that my SEE is still not totally tested, so it might have some bug.
José C.
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.