Author: Robert Hyatt
Date: 13:00:22 09/25/98
Go up one level in this thread
On September 24, 1998 at 21:51:55, John Coffey wrote: >When we talk about move ordering, are we talking about sorting all the moves >or just trying to pick the best one? I think that we may covered this before, >where Robert said .... > >"generally accepted terminology says that if, when you get a fail-high, >it happens on the first move at least 90% of the time, you have what is >called a "strongly-ordered game tree with near-perfect ordering." Crafty >is sitting at around 92-94% on this statistic..." > >John Coffey the only thing that matters is picking the best move first, when you can get a cutoff. At every other ply in a normal search (not on the PV move however, which is more complex, but on every other root-move) you will find that at ply=2 you expect to get a cutoff on move 1, while at ply=3 you don't and have to search all the moves, but at ply=4 you get a cutoff on move one, and this repeats, assuming your move ordering is perfect. So getting that best move first is important on nodes where you expect a cutoff. However, if the best doesn't cause a cutoff, they you try another good one assuming your "best" really wasn't best. After a few of these, you might become convinced that this is a non-cutoff node and give up on the ordering. I try the hash move, winning/even captures, 2 killers, and then 4 history moves. If I haven't gotten a cutoff by here I quit fooling around and search the rest in generated order...
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.