Keith123 wrote:
Wrote:

According to the endgame tablebases: Position=0, Kf1=0, Kd2=0, Ke2=0,
Kf2=0, e5=0, Kd1=0. In other words, all moves draw and are silly!

Thanks for the reply. Just seems weird that Chessmaster would send a
pawn on a guaranteed suicide mission when White might have a chance to
win if Black does not play perfect. It was easy for me to see not to
follow the technique of Chessmaster in this simple endgame test, but
what other situations might lead me the wrong way in my learning?! I
will try some of your "tweaking" suggestions and see how it goes. Oh
well, right now I can't reach an endgame in a real game against
Chessmaster anyway!

Almost all implementations of endgame databases will play a
"random-looking" move when the game is determined to be a draw. Another
example is KRB vs KR, when the computer will trade rooks rather than
force the opponent with less material to "prove the draw".

Only a very few engines have implemented what is called a "swindle
mode" for endgames. Crafty is the only one I know of for sure, and it
typically only uses it when it knows it's playing against a human. In
this mode, it will eliminate the kinds of moves that you saw against
Chessmaster in the hopes that the opponent won't properly play the
endgame.

Here's how Chessmaster would play out the ending without endgame
databases:

[Event ""]
[Site ""]
[Date "2006.8.28"]
[Round ""]
[White "CM NO EGDB"]
[Black "CM NO EGDB"]
[TimeControl "?"]
[Result "1/2-1/2"]
[Setup "1"]
[FEN "4k3/8/8/8/4P3/8/8/4K3 w - - 0 1"]

1.Kd2 Kd7 2.Ke3 Ke6 3.Kf4 Kf6 4.Kg3 Ke5 5.Kf3 Kd6 6.Ke2 Ke5 7.Kd3 Kf6
8.Kd4 Ke6 9.e5 Kf7 10.Ke3 Ke6 11.Kf4 Kf7 12.Kf5 Ke7 13.e6 Ke8 14.Ke4
Ke7
15.Kd5 Ke8 16.Kd4 Kf8 17.Ke5 Ke7 18.Kd4 Kxe6 1/2-1/2

jm