AXchess 0.8 and search windows ...

[Brice Boissel]

Geschrieben von: / Posted by: Brice Boissel at 28 July 2002 18:52:23:
AXchess is nearly perfect.
But : why the search windows is including only 2 half-moves in the search variantes ?

[Pham Hong Nguyen]

Geschrieben von: / Posted by: Pham Hong Nguyen at 29 July 2002 00:55:58:
Als Antwort auf: / In reply to: AXchess 0.8 and search windows ... geschrieben von: / posted by: Brice Boissel at 28 July 2002 18:52:23:

AXchess is nearly perfect.
But : why the search windows is including only 2 half-moves in the search variantes ?

Hi,
I have made some changes in pv code and search results seem be little better. I agree that display of pv length is much shorter by unknew reasons (Ax checks and prints out only legal moves of pv). I will pay more attention on it. Thanks for your feedback.
Regards,
PHN

http://www.geocities.com/axchess

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 29 July 2002 08:03:26:
Als Antwort auf: / In reply to: Re: AXchess 0.8 and search windows ... geschrieben von: / posted by: Pham Hong Nguyen at 29 July 2002 00:55:58:

AXchess is nearly perfect.
But : why the search windows is including only 2 half-moves in the search variantes ?

Hi,
I have made some changes in pv code and search results seem be little better. I agree that display of pv length is much shorter by unknew reasons (Ax checks and prints out only legal moves of pv). I will pay more attention on it. Thanks for your feedback.
Regards,
PHN

I played some fast games with it against latest movei at
10 seconds per game and Ax lost on time(both movei and ax used
16 Mbytes hash and the hardware was p800).
It even lost on time in 1 minute/40 moves and the shortest
time control that it could play without losing on time
is 2 minutes/40 moves.
Ax beated movei 5.5-2.5 at that time control
(Movei always changed the first move
and it means that it can starts with bad moves like 1.a4
but the same movei could beat amateur at that conditions at
1 minutes/40 moves in 40 game match).
5.5-2.5 is not a significant result and
in the match against amateur the result was
7.5-2.5 for amateur after 10 games and
12-8 after 20 games but movei finished the match
by winning 21.5-18.5 if I remember correctly.
The same movei beated movei0.07a in 40 game match
at 15 minute per game 27.5-12.5.
I have positive results also at shorter time control and based on report of peter berger older movei is not without chances against yace
(of course yace is clearly better but 4-1 for yace with 3 draws
is a very good result if you consider the fact that
movei knows almost nothing about pawn structure
and nothing about king safety,and not much about
using hash tables).
I plan to release a new version only when I fix part of
my hash problems and fix a problem of wrong pv in the qsearch
and improve my qsearch and I do not know
when I am going to do it.
I do not plan to add knowledge about pawn structure
and king safety in the next public version(there is
an improvement in the evaluation but the main change
is that movei is less materialistic).

I try first to use hash tables to avoid null move pruning and
later to use them to avoid generating the legal moves but
I do one change and test and
I still did not succeed in using hash tables to avoid
null move pruning correctly.
Uri