[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Suggestion for change

From: Tim Bedding
Subject: Re: Suggestion for change
Date: Thu, 31 Oct 2002 11:30:22 -0000

Simon wrote
> Given the position with 40 moves in 5 minutes the test release
> of 5.05 plays Qe5 at about 40 seconds on this hardware (P200
> equivalent), as you suggest.

This is surprising. I would expect you to see the same
moves as me.

I just checked
and it seems that the last-changed date is yesterday.

My guess is that you are seeing different play because
of different evaluation code.

Is the stronger evaluation code you mentioned to me
recently in your new download file?

Do you get Qe7 as the following move?

> Evaluate already has a simplistic trade down bonus, might be
> scope for tuning it, but you'd want several endgame positions.

> I suspect that specifying ply depth might be having an adverse
> approach to this position, stick with time.

I think I was using time and I got Be5 from the debug
output for various output levels.

> There are also some tactical possibilities in this position for
> white to gain material, which I'd expect a computer might prefer
> given enough time, as given a chance those pawns might fall with
> check, that would be a typical computer approach to endgame
> tactics, and isn't objectively worse (if you are a computer and
> don't blunder badly) I think.

Well, I was reasoning as follows and would like to know
what you think.

I played GNU Chess against Crafty (I think) from the position
I gave and saw that the position seemed to get bogged down
in a lot of queen moves without much progress towards

Exchanging queens cut through all that, eliminating
the possibility of perpetual check which will force
a draw.

The penalty I had in mind would be as small as possible
and thus would be outweighed by a route leading to
material gain.

> One more obvious area for "evaluate" going wrong, is I believe
> its insistence on not moving pawns in front of the king, even
> when back row mate threats start becoming a significant part of
> the analysis. I note in this position no pawns have been moved
> in front of the computers king! Whilst not moving them lightly
> is good chess, GNU is a tad dogmatic on these things!

An interesting point but I cannot see offhand how I could
come up with a formula for that one.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]