Author: Daniel Clausen
Date: 01:53:25 11/11/02
Go up one level in this thread
On November 10, 2002 at 23:51:13, Robert Hyatt wrote: >On November 10, 2002 at 22:41:06, Russell Reagan wrote: > >>On November 10, 2002 at 21:29:43, Robert Hyatt wrote: >> >>>This means that 1/2 of the time, you will >>>predict correctly and when your opponent moves, you have an instant response >>>ready. >> >>Have you ever seen instances where one engine dominated another by always >>pondering the correct move? I wonder how valuable this would be. For example one >>example of why this might not be THAT effective, if you are essentially able to >>search for twice as long (meaning you have double time odds), that's only worth >>one extra ply if you're effective branching factor is 2, right? So most of the >>time you won't even get an extra ply, and at the depths reached at tournament >>depths, an extra ply might not even be decisive IF you were able to reach it. >>So, how valuable is guessing the opponents move in practical play? >> >>Russell > > >This is pretty common. What typically happens is that program A goes out of >book first and does a normal search. Program B is doing nothing. A makes a >move and starts pondering the right move. B starts searching. Every time B >makes a move, A has been pondering the right thing and makes an instant >move. But it is rare to predict 100%. So eventually A predicts the wrong >move, B makes a move and now A has to think, while B ponders the right move. > >This kind of swing happens repeatedly during a game... So it can even be an advantage to be out of book first. :) One possibility could be to do a little search after a few moves (let's say from the 5th move) but still play the book move if the search doesn't find anything interesting. Of course you'd need additional book information then, like 'don't play b5 here!' because otherwise your search would jump into all sorts of traps. On the other hand, you could prevent doing a major fault in the opening, which could be useful for books created from a bunch of PGN-files, w/o verifying their correctness. Sargon
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.