Author: Robert Hyatt
Date: 07:41:01 11/24/02
Go up one level in this thread
On November 23, 2002 at 08:11:37, scott farrell wrote: >Just after other people's thoughts. > >I think Omid's work overlooked the adapative null move searching many of us do, >ie. transitioning from r=3 to r=2. > >I think adaptive null move tries to GUESS where to use r=2 to reduce the errors >that R=3 makes. I guess it depends on how often this GUESS is correct, the cost >of the verification search, and how long it takes the adaptive searching to >catch the error at the next ply. > >Has anyone looked at setting the verification search to reduced depth of 2 >(rather than 1)? obviously to reduce the cost of the verification search. > >Robert : How did your crafty implmentation go? I have not had time to finish it yet. We've been ordering a bunch of new machines, my ftp machine developed a disk error, a lab was broken into and some minor equipment stolen, etc. Hopefully I will get a chance to work on it some more tonight. > >I know that measuring nodes for a fixed search depth sounds like good science, >but I think you really need to look at where one method makes a mistake, and how >quickly it can find the mistake at the next ply. Obviously my point is if it >finds a given move with less nodes, thats good, if it takes an extra ply of >search, sometimes that's acceptable if we can catch it quick enough. We've all >seen how quickly at the next ply a fail low is often fast, well that is my major >"verification search" for ALL search problems - horizon, null move, etc etc. I >think it might be better to include some nodes from the next iteration, where >the next ply fails low really really quickly, and finds the mistake anyway, >maybe consider the nodes as part of the previous ply. > >Scott Whether it works or not is probably going to depend on the program. Which is not unexpected. I think that most of the null-move errors happen right at the border where the search collapses to a q-search only. This might help some of them (but not the ones where R-1 also collapses to a q-search only.) And it will obviously handle the cases where zugzwang plays a big role, such as in endgames, which might allow null-move to be turned on always...
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.