Author: Robert Hyatt
Date: 08:21:00 09/03/02
Go up one level in this thread
On September 03, 2002 at 10:36:58, Uri Blass wrote: >On September 03, 2002 at 10:22:46, Robert Hyatt wrote: > >>On September 03, 2002 at 06:58:41, Gian-Carlo Pascutto wrote: >> >>>On September 02, 2002 at 19:33:48, Robert Hyatt wrote: >>> >>>>Near the tips of the tree? traditional killers, captures, whatever, works >>>>quite well out there... about as well as hashing, since the depths will be >>>>so shallow... >>> >>>If you do them, yes. But IIRC, the DB hardware had no killers, just >>>MVV/LVA captures. >>> >>>-- >>>GCP >> >> >>I actually don't remember whether it does or not, but I suspect not. But >>captures are _the_ most common refutation move anyway. I would never argue >>that their alpha/beta hardware search is as good as a software alpha/beta >>search, for lots of reasons. But if all you are doing 99.9% of the time, >>is just proving that a score is >X or <X, then no re-searches are needed, and >>efficiency is not an issue whatsoever with respect to hashing. > >Hasing is clearly important to prove that the score is >x or <x and the same for >killers. > >Without them you may often start with the wrong move in your search. > >Uri this has nothing to do with mtd(f). In the discussion at hand, you search a shallow tree of 4-5-6 plies and prove the score is >X or <X. You don't re-search it a second time as a mtd(f) search would, so hashing has no real significance. We are therefore talking about the _first_ time we search a position, and hashing will have zero information to help. If we have to search it a second time, as with a normal mtd(f) algorithm, then hashing may help (if we are still on the same wrong side of the true score) or it won't help at all if we are on the other side of the true score. (I'll leave it as an exercise to the reader to figure out why this is easily provable as true.)
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.