Author: Djordje Vidanovic
Date: 12:02:06 12/13/00
Go up one level in this thread
On December 13, 2000 at 14:58:34, Djordje Vidanovic wrote: >On December 13, 2000 at 14:15:06, martin fierz wrote: > >>hi, >> >>after reading the description of mtd(f) it seemed to me that this kind of >>algorithm could only work if the whole tree still fits in the hashtable, and >>that once it doesnt fit in any more it would get terribly inefficient. is this >>right or wrong? are there any chess programs using mtd(f)? i once read that the >>most popular search is PVS where the first move gets searched normally and all >>others with a minimal window - is this correct? >> >>cheers >> martin > > >In MTD (f) positions are tested to see whether they are higher or lower than the >guessing (done successively for each position). Memory is hash tables, while >the Driver tries to assemble the scores for the given position. As the (f) >designates the first "guess", it should be rather accurate (using the previous >scores), so one can assume that the tree should get overly large >(supposedly...). > >Apart from AnMon by Christian Barreteau that is an MTD (f) program, there is >also PostModernist by Andrew Williams that plays on ICC quite successfully. For >the time being, PM runs only under Linux and is still a private program. > >*** Djordje >>>tree should get overly large (supposedly...). should read as ..."the tree should NOT get overly large..."... but I guess it was clear from the previous clause. *** Djordje
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.