Author: Gian-Carlo Pascutto
Date: 15:04:28 12/27/00
Go up one level in this thread
On December 27, 2000 at 17:27:59, Vincent Diepeveen wrote: >Sjeng: prove > >You typed in PROVE and it then shows: "position is won, without >showing any 'mate in xx' or whatever. > >So if i confused you with a mate prover, i'm sorry for that! It's called 'prove' because the algorithm it uses is named proof-number search...but it isn't a mate prover, as there's no guarantee it'll find the shortest mate. >>They no longer overextend forced checking sequences. You still do. > >This is the biggest nonsense i ever heart. >You tell me, having made a program which uses less nodes a ply as >any other program that's not forward pruning other as nullmove, >that i'm 'overextending'? If that is indeed true, well then congratulations. At how many ply do you see the 60 ply solution? >Perhaps you should rephrase that! This is more like a difference I view...I have my doubts about whether it makes sense to see 60 ply sequences in, say, a 10 ply search. If you can do that however and still produce smaller trees than anyone else well then hats off. Restricting forward pruning in your comparisation to nullmove is too strict though. Any forward pruning technique which does not make the program noticaly weaker is fine by my book. >>Oh and by-the-way...Sjeng without the matefinder finds the correct >>move in +-170 nodes. > >This is complete bullshit unless you have a lot of luck with >move ordering which is one in a million or something here, >or you count nodes in a very weird way. Reread what I wrote...Sjeng finds the correct MOVE in 170 nodes ;) It doesn't give a mate score in the first 14 plys...but in a game it would play the correct move and that is all that matters I think. -- GCP
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.