Author: Vasik Rajlich
Date: 02:22:26 06/16/05
Go up one level in this thread
On June 15, 2005 at 22:40:22, Tor Alexander Lattimore wrote:
>Hi
>Is it possible to use a single variable in the MTD(f) search? Something like
>this:
>
>int MTD(int depth, int guess)
>{
> if (depth<1) return Evaluate();
> MOVE move_to_search;
> int best=-INFINITY;
> GenMoves();
> while (GetNextMove(&move_to_search))
> {
> PlayMove(move_to_search);
> val = -MTD(depth - 1, -guess + 1);
> UnPlayMove(move_to_search);
> if (val>best)
> {
> best=val;
> if (val>=guess)
You missed a line here. Fail-hard is "return guess", fail-soft is "return val".
Fail soft helps when you need to re-search, so it helps more in MTD (f) than
PVS, and doesn't matter at all in pure alpha-beta.
Vas
> }
> }
> return (best);
>}
>
>perhaps there is something very wrong with this? or perhaps it's used already, I
>just noticed that on Aske Plaat's site he always uses an Alpha-Beta search with
>0 width windowed searches, but doesn't this do the same thing? Is using
>fail-soft type algorithms used in MTD(f) since it could well help zoom into the
>correct score sooner?
>
>Cheers
>Tor
This page took 0.03 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.