Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame stupidity

Author: Andrew Williams

Date: 04:46:09 11/29/00

Go up one level in this thread


On November 29, 2000 at 04:14:30, Scott Gasch wrote:

>Hi all,
>
>I've been seeing my engine make some really dumb decisions in endgames receltly.
> It just can't seem to see deep enough to decide to do the right thing.  Here's
>an example position:
>
>[D]8/p5k1/3p2p1/2pP1p1p/8/7P/6P1/3B3K w - f6 0 36
>
>This happened in a game where it was playing black.  Not good.  But the way it
>handles it is even worse.  It lets the white king walk around and kill the A
>pawn then the two in the middle realizing too late that it could have tried to
>defend them with its own king.  Likewise when I have it play white it does not
>see this sequqnce opting instead to trade pawns, pick off a couple with the
>bishop and just plain mess everything up.  It never moves the king and does not
>realize it's hosed until here:
>
>[D]8/8/3p2B1/8/p1pk4/8/8/7K w - - 0 0
>
>I obviously need help... a different eval routine specifically designed for
>endgames maybe?  I know my search is sound, I am just not getting deep enough
>fast enough to see endgame tactics and getting routinely punished for it.  To
>try to improve things I added code giving large bonuses for passed pawns and a
>pawn-push extension... it seems that this has not helped matters though.
>
>Thanks again,
>Scott


With hardly any pieces on the board, you want your king involved more. There's
no need for him to hide in the corner. There are lots of things I tell mine,
for example:

1 Go to the centre.
2 Support my passed pawns.
3 Block my opponent's passed pawns.

However:

(a) It's difficult to decide which of 2 and 3 to do in a particular position.
(b) Sometimes the centre is the worst place to go
etc etc

Andrew



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.