Author: Vincent Diepeveen
Date: 18:11:20 07/18/99
Go up one level in this thread
On July 18, 1999 at 17:14:17, Dave Gomboc wrote: >On July 18, 1999 at 14:22:14, Vincent Diepeveen wrote: > >>On July 18, 1999 at 14:10:03, Andrew Williams wrote: >> >>>In PostModernist's case it would search 1.03, then 1.02, then 0.98, then >>>0.89, then 0.73, then 0.48. It would then start to work its way back up, >>>0.48, 0.49, 0.51, 0.54 etc. >> >>>So in this example, PM needs a lot of re-searches, but it's not just to >>>do with the gap between the iteration scores - for example, going from 1.03 >>>down to 0.48 would be considerably faster than going from 1.03 to 0.72, >>>because of where the scores lie and because of my approach to traversing >>>gaps. >> >>The idea of having more than 1 research is laughable. >>Now you can of course make a statement that the overhead >>isn't *that* high, which is more or less true, but still... >You're forgetting that the searches all cost far less than a PVS search. You >can sum the total time of 6 researches compared with a PVS implementation, and >the mtd(f) will still be faster more often than not. That's not true, as in the worst case your window is a pawn off. So the first 5 researches are searching space which is completely useless, where in PVS the search overhead only depends upon move ordering. >If you tried mtd(f) with a serious effort and could not get it to work better >than horribly, my guess would be that the stepping algorithm you used for >researches probably was not up to the task. If you have to step a lot, which is logical as my evaluation ain't dumb, then it's quite logical that you search more nonsense, which makes it perform bad.
This page took 0.01 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.