Pawn-ending extension

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
hgm
Posts: 27790
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Pawn-ending extension

Post by hgm »

I just implemented an extension of 50% of the remaining depth for moves that convert to a Pawn ending in KingSlayer. This to prevent it from converting to such a lost end-game by unnecessarily trading its last piece because it lacks the depth to recognize it as lost. While it could have added that depth with negligible impact on the total tree size, as the branching ratio in Pawn endings is very low. So that is what it does now.

The results are a bit funny, though. When it now is ahead one-to-nothing in pieces, it often just gives away the piece for free, to get the extension. Because only with the extension the promotion is within the horizon. This has no effect on the result, of course; it wins those games anyway. But it looks a bit silly...
JVMerlino
Posts: 1357
Joined: Wed Mar 08, 2006 10:15 pm
Location: San Francisco, California

Re: Pawn-ending extension

Post by JVMerlino »

hgm wrote: Wed May 29, 2019 10:50 pm But it looks a bit silly...
No sillier than some endgames (or even conversion to endgames) from engines using endgame tablebases....
Volker Annuss
Posts: 180
Joined: Mon Sep 03, 2007 9:15 am

Re: Pawn-ending extension

Post by Volker Annuss »

That's easy to avoid when you extend recaptures only.