Author: Robert Hyatt
Date: 07:43:10 11/14/01
Go up one level in this thread
On November 13, 2001 at 11:43:28, Uri Blass wrote: >On November 13, 2001 at 10:33:43, Dieter Buerssner wrote: > >>On November 13, 2001 at 04:22:02, Uri Blass wrote: >> >>>I believe that yace with the existing tablebases can find the right move in >>>almost every KPPP vs KPPP endgame by search in a short time. >> >>Uri, thanks for your confidence. However I fear, this is unfortunately not true. > >It depends on the definition of almost >We can try to generate 1000 random legal KPPP vs KPPP endgames and find in how >many cases programs can find the right move. > >The only possible problem is that in some of these positions humans may also not >know the right move. > >Uri The problem is in _this_ position, there is only one right move for each ply (for white to maintain the forced win). That causes problems because the branching factor is non-trivial since in many lines pawns promote, which causes great problems. And when you get to the right positions, promotions are forced... The neat thing is that white can't make a single wrong move without turning this into a draw or loss. And once white makes a wrong move, then black can't make a single wrong move without doing the same thing to his winning chances. It is a _very_ deep, _very_ narrow "proof" of wtm wins in every case.
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.