Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Can your prog solve this -- results

Author: Robert Hyatt

Date: 13:43:51 02/18/99

Go up one level in this thread


On February 18, 1999 at 16:30:01, Will Singleton wrote:

>
>Here are the results:
>
>                    <=5    6     7     8     >=9
>
>Fritz 5.32          dxe5   dxe5  dxe5  dxe5  dxe5
>Hiarcs 6            Nh4    Nh4   Be3   Be3   Be3
>MacChess 4          dxe5   dxe5  dxe5  dxe5  Bc1
>Crafty 16.5         dxe5   Bc1   Bc1   Bc1   Bc1
>Amateur 0.91        dxe5   dxe5  dxe5  dxe5  Bc1
>PostModernist 0.9   Nh4    dxe5  dxe5  Bc1   Bc1
>EXchess 2.5x        Bc1    Bd2   Bc1   Be3   Be3
>LambChop 7.x        Bc1    dxe5  Bc1   Bc1   Bc1
>Gromit 1.2?         ---    ---   ---   dxe5  Nh4
>Robertson 0.0       Bc1    Bc1   dxe5  dxe5  dxe5
>Ferret              Bc1 or dxe5, whatever
>InmiChess           ---    ---   Nh4   Nh4   Nh4
>Diep                classified (private email :), but it wasn't dxe5)
>
>
>Thanks for all your responses, we have some interesting results.
>A couple of issues here -- the first is that the position
>occurred in a blitz game, so only about 10 seconds was available
>to move.  As it turns out, the position is already bad, -1 or
>so (as Bruce says, White is hosed), so there's no real good move
>-- but short term, there is a wide variety of opinion between
>the various programs.  This has significance for blitz play,
>especially considering that dxe5 is probably the best move!
>
>I thought it might be interesting to analyze the reasons for the
>diversity of move selection.  Some programs see deeper for a
>given ply depth, of course.  Do they pay a price for this by
>having to take longer to iterate to a given ply?  Or does the
>deeper search (more extensions) result in better ordering, which
>causes the desired ply to be reached even sooner?
>
>The position is fairly complex, and I would expect lots of
>extension triggers.  For example, Hiarcs is known (I think) to
>extend quite a bit.  I ran this with Hiarcs 6, and it never
>considers dxe5 (as the best move) through many minutes of
>searching.  But after dxe5 a4 Qxa4 fxe5 Qxc4 etc, it has a
>better eval for white than it did with any alternative to dxe5
>originally.  So it looks like it missed what it later considers
>to be the best move, which is dxe5!
>
>Bc1 is worse (or no better) because of a4 Qb4 Re6 Nh4 Rb6 ...
>Be3 is worse (or no better) because of exd4 cxd4 Re6 Re3 Rb6 ...
>
>Maybe Bob has a bug. :)  In both outputs I received, Crafty
>doesn't see the proper continuation after dxe5.  And I believe
>Dann said that it may have switched back after a long think.
>

I don't think I have a 'bug' but we may have differences...  IE Dan wasn't
using the current 16.6 code, which is significantly different in the eval
area.  On the above test (the data I sent came from my notebook) it liked
Bc1 until the last ply I gave you when it changed to Be3 and stuck there.

I am not sure de loses...  you may be lost already...




>Interestingly, Jim Walker says that Fritz 5.32 prefers dxe5
>throughout a long search.  I would like to see the pv output on
>that.  Why does Fritz remain constant on what is probably the
>best move, while others are all over the lot?
>
>Will



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.