Author: Robert Hyatt
Date: 18:48:15 12/25/98
Go up one level in this thread
On December 25, 1998 at 06:14:47, blass uri wrote: > >On December 24, 1998 at 19:03:00, Robert Hyatt wrote: > >>On December 24, 1998 at 11:02:09, blass uri wrote: >> >>> >>>On December 24, 1998 at 10:52:40, blass uri wrote: >>> >>>> >>>>Suppose you do a brute force search to depth x then what is the longest lines >>>>that you may search? >>> >>>I mean seeing everything for x plies when I say depth x. >>> >>>Uri >>> >>> >>> >>> >>> >>>> >>>>I understood that for Junior the longest line is at most 4x and for crafty the >>>>longest line is at most 2x. >>>> >>>>correct me if I am wrong >>>> >>>>What is the longest line for other engines >> >>first you have to have a common "depth" term. The depth junior displays is >>_not_ the same depth the rest of us display, so you can't compare that. > >I understand and this is the reason that I explained that I mean seeing >everything for x plies(it should be almost everything because Junior cannot see >underpromorions so it cannot be everything but in more than 99% of the cases the >result is the same as the result when it considers everything). > >If Junior displays depth 2x then I say because of this reason that the depth is >only x. > >> >>But in the case of Crafty, the nominal depth of search has nothing to do with >>the number of moves in the PV. IE with a 10 ply search I can hit ply=50 at >>times if the extensions get triggered just right. Someone posted a position >>months ago that was a mate in 30, that Crafty found with a fairly shallow >>(10 plies or so) search, yet it went a full 60 plies deep. This position was >>one of the "rebel" positions where there was a contest to construct the deepest >>mate with the fewest nodes... > >The question is what is the maximal plies that you can hit with a 10 ply search >if you do not consider forced moves(only one legal move) and replies to forced >moves. > My max ply limit is 60, period, although anyone can recompile this and make it larger (it will slow things down a percent or so to do so). But with the check extensions, recapture extensions, pawn push extensions and threat extensions, 60 plies can be hit with no trouble... >> >>But when you see 10 plies, you do know that the pv was searched to at least 10 >>plies, but could possibly have went to 50 plies (or more) along some variations. >>Also you know that other non-PV moves may well have not even been searched to >>10 plies thanks to the null-move... > >The question is what is the probability in practical games that you miss >something important because of the null move. > >In other words suppose that you are not using null moves then what is the >probabilty that you do a different move with a 10 ply search? > >Uri fairly high in some positions, not at all in others. It is just a form or forward pruning... and it affects the score and/or PV on many occasions...
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.