Author: Dann Corbit
Date: 10:47:21 01/07/00
Go up one level in this thread
On January 07, 2000 at 13:43:09, Ian Osgood wrote: >Here is a little jaunt away from the same old alpha-beta/tablebase discussions: > >There was a recent posting about a helpmate problem composed by Nunn which >stumped Karpov & Kasparov. The problem: White has moved 1.e4. Come up with a >line which results in 5.NxR checkmate. You must determine whether this is >White's or Black's fifth move. > >I have been thinking about how to solve this problem on a computer. Since it is >a help mate with a very specific end condition, I would think a standard >alpha-beta search with a positional eval wouldn't work. First of all, the eval >or search should detect the end condition at the appropriate plies. Second, the >search should be full width, or prune based on problem specific info (for >instance, can a knight reach the king to deliver the mate on time). I would >think a hash table would help to speed up any full width search. Since we are >searching for a specific condition, perhaps a different search method such as >proof-number search would be more appropriate. > >Does anyone have any experience writing helpmate solvers? Anyone have any other >ideas about solving Nunn's problem? Also, I am looking for public source code >for Proof-Number searching. Heiner Marxen's Chest program also solves helpmates (and many other sorts of chess problems as well). Comes with source. Worth a look. UNIX version: http://www.drb.insel.de/~heiner/Chess/chest.html Win32 version: ftp://38.168.214.175/pub/chest/ It is (by an overwhelming landslide) the best chess problem solver on earth.
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.