Author: Dave Gomboc
Date: 19:20:48 09/22/98
Go up one level in this thread
On September 22, 1998 at 12:17:03, Pat King wrote: >Well, upon pondering the problem I made a refined estimate (but >still only an estimate) of 2^32 or about 4.3 e+09 nodes, which >works out to 20.2 plies instead of 19.something. Is the 2^32 >figure for P(lookup error)=.5 for a 32 bit key a coincidence? I >suspect not. If not, then you can easily decide on how much error >you're willing to tolerate. Based on this analysis, 64 bits >certainly seems like overkill to me. > >Pat We know from experimentation that 32-bit hash keys are not enough to prevent a fatally high number of collisions in a 3-minute search. The next logical number ot use is 64 bits, simply because it is the next higher power of two. Sure, maybe you could try it with 48 bits or 56 bits, but why bother? It's going to cost you more time than it is worth for the extra space that you save in the hash table. Dave Gomboc
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.