Author: chandler yergin
Date: 21:07:03 12/03/05
Go up one level in this thread
On December 03, 2005 at 18:53:20, Roger Brown wrote: >>>Mate in 11? Not my Shredder 8! >>It found a Mate. Does this mean it was the best line? No, of course not. >>As you can see with deeper search it reverted to Mate in 14. >>How often do you see a PV Change with time? >>This is no different. > > >Hello Chandler, > >Your statement is difficult for me to understand. > >When I use Franz Huber's wonderful Chest UCI and it tells me mate in 5 it means >that whatever the opponent replies with, mate is going to come in that number of >moves. > >Let me use a simple example. > >Mate in one means that after the mating party moves, there is no possible escape >for the enemy king. > >For an engine to say that there exists a mate in 11 then say a mate in 14 for >the same position means that there is a problem. > >A mate in one is a mate in one. An engine that subsequently evaluated it as a >mate in ten has a problem with mate calculations. > >The other way round, finding a mate in ten which, when tested with a matesolver >such as Chest, really is a mate in eight, just means that the engine has found a >less efficient solution that still leads to a mate. > >Do you understand now why there is something wrong with your statements about >depth and time? > >Later. How often have you seen a PV show Mate in 150? Deep Search maybe 18/45 Ply. It doesn't show a forcing line of course. With time the PV evtually finds Mate in lesser moves,and then with a Fail High it changes rapidly, and maybe at Ply 35/65 can demonstrate the Mate. Because the Program sees a Mate after a 1/2 move is made from the Mate in 14, means that the line being evaluated changed slightly allowing a Mate in 11. The next 1/2 move = 1 Ply the best line again shows Mate in 14. No one expects the Program to show the forcing line at the time it finds the Mate in 14. The Engine would have to search to 28 Ply to do that. Programs evaluate 1 Ply at a time. Hash Tables are not perfect, if so, Computers would 'never' make blunders, yet in games, we see it all the time. They just plain make 'bad' moves at critical times. Why is that? Because they make the best move found.. at the time the software tells it to move. How often have you seen a Program announce Mate in 12 for example, but when you force the play it turns out to be 15 or more? Top Programs are not the best Mate Finders! Simple as that.. OK?
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.