Author: Robert Hyatt
Date: 20:37:35 11/30/02
Go up one level in this thread
On November 30, 2002 at 11:02:53, Uri Blass wrote: >On November 30, 2002 at 10:39:45, Robert Hyatt wrote: > >>On November 30, 2002 at 09:27:20, scott farrell wrote: >> >>>This is a continuration from another thread, but I need some help in >>>understanding this. >>> >>>Some of the engines "failed-high" detecting the move, before they had fully >>>completed the ply. >>> >>>Uri and I cant understand how they detect or report this. >> >> >>Simple. When you start an iteration, you should use some sort of aspiration >>window, rather than setting alpha=-infinity and beta=+infinity. In the case >>of Crafty, if the score for iteration N-1 was +.30, then I might start the >>search at iteration N with alpha=0, beta=+60 to try to bracket the expected >>score. This makes the search tree smaller as you can prune lines that lose >>or win material quickly. However, if you really do suddenly see a path to >>win a pawn, that score will be >= +.60, which means the search returns beta >>rather than a true score. You have to re-set beta to something larger and >>search again to get the true score. That is a "fail high". > >You missed the question(that I already deleted >in this post not to do it too long because I repeat it) > >The question of the poster was not what is fail high >but how programs detect fail high for a move >before getting a score. > OK... the most plausible explanation there is that the first move (or moves) all produced scores <= alpha, so no PV was produced. Most seem to do as I do and if the first fails low, I display that with a fail-low notation and then re-search it to see how bad it is, before going on. But it is also perfectly legitimate to continue searching and you might, on occasion evan fail high on another move... Which way is better is debatable. But for a fail low, I want to know the score so I can figure out how much time to use searching for something better... >The program finished depth 10 with Nf6 and >suddenly wanted to play Nf4 at depth 11 without >getting a score for Nf6. > >I guess that the reason is that it did >a search with a window (+3.76,+3.77) >to find if there is something tactical >before doing normal search with a window of >(+2.3,+3.76) in case that it does not find >something that is very good. > >The advantage for test positions is that >you can often solve them faster. > >The disadvantage is that in game you usually >waste more time because there is >no big fail high. > >It may be a good idea to do it in part >of the cases if you have a way to predict >with big probabilty that a move may >fail high in the next search. > >I believe that doing it always is a bad idea >but I never tried it. > >Uri
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.