Author: Don Dailey
Date: 08:13:57 01/18/98
Go up one level in this thread
Christophe, I think this is how most people do it, I was just trying to "sync" with everone else. I think the simplest way to state it is everytime your progam makes a move. I don't count null moves but I have a feeling it's very minor. Whenever there is some doubt I choose the conservative way becasue I want to know my counts are at least a lower bound when comparing to others. But it sounds like most do it the "right" way. About counting illegal moves. Bob argues these should be counted. I have no problem with this. I said in an earlier post this was inefficient but should not have. I once did this too and it was indeed a speedup but took it out for some technical reason having to do with the parallel code I believe. - Don On January 18, 1998 at 01:51:03, Christophe Theron wrote: >On January 17, 1998 at 16:43:39, Don Dailey wrote: > >I count the number of positions my program really sees in its internal >board. > >Very simple: I have a counter which initial value is 1 (the root >position is counted). Then it is incremented at each call to my >"makemove" procedure. A null move counts as a move, because I really >call "makemove" for a null move (the hash key has to be updated, this is >the job of "makemove"). > >This is not equivalent to counting the number of calls to search() and >qsearch(), because I don't call them when I see the position is not >legal. > >I do generate, and make/unmake, illegal moves. Two days ago I measured >how much, because I was wondering if trying not to generate them could >efficient. They represent roughly 2% of the nodes counted by my search >(could be different for yours), so I decided that any attempt to avoid >them was likely to slow down Tiger. > > > Christophe
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.