Author: Robert Hyatt
Date: 16:25:56 09/16/00
Go up one level in this thread
On September 16, 2000 at 15:36:22, Larry Griffiths wrote: >On September 16, 2000 at 14:30:16, Robert Hyatt wrote: > >> >>You have found the fool-proof way of preventing collisions. However, you >>also see the cost (that of storing a real board position, even if compressed >>a bit). >> >>Bottom line: some errors are acceptable. Searches are going so fast, the >>probability of a collision is pretty high, but the probability of it happening >>on a node that is "important" is very low. Which means acceptable. > >Thanks Robert! > >I needed verification that this happens with hashkeys. > >I also noticed that the lower parts of my hashtable have less >holes in them when I use hashtable sizes that are not powers of 2. > >Larry. This last thing suggests that maybe your hash keys are not as random as they should be. I have done some hash testing and I don't see any "holes". IE I can quickly write over nearly *every* position. I detect this by simply running a search, then counting how many entries are from this search, vs previous searches. I can get 99% easily...
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.