Author: Robert Hyatt
Date: 13:17:06 01/14/00
Go up one level in this thread
On January 14, 2000 at 13:48:09, Landon Rabern wrote: >On January 14, 2000 at 00:45:25, Robert Hyatt wrote: > >>On January 13, 2000 at 18:34:12, Landon Rabern wrote: >> >>>Is it known what information is sufficient to garantee that a hash entry is >>>describing the same board as another board hashed to the same value. Like can I >>>store a 64 bit int with all the pieces on the board and use this to test? Or >>>maybe I can use this 64-bit int and test the the from-to of the suggested move >>>has the right pieces on them? >>> >>>Thanks, >>> >>>Landon W. Rabern >> >> >>To "guarantee" a match, you need roughly 160 bits... whatever it takes to >>_exactly_ encode the board position. :) > >What I actually meant was how many bits I needed in my key to check if it was a >collision. Or is this what you meant? I am using 64-bits right now and >checking the squares in the suggested move to see if they work. It doesn't seem >to mess up anywhere. Will it maybe corrupt the board sometime? > > >Thanks, > >Landon If you take the hash move and make it without being certain it is legal, yes it will come down around your ears. If you validity check the move, which is what Bruce meant when he said to "make sure you don't crash if you get a false match" then you will be ok. 64 bits will hardly get any errors, but "hardly" and "never" aren't the same. Just be sure you can absorb errors without crashing, because the errors probably will never affect the root score.
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.