Author: Dieter Buerssner
Date: 13:31:15 12/05/01
Go up one level in this thread
On December 05, 2001 at 15:46:12, David Rasmussen wrote: [...] >>sequence with a good Hamming distance in the cited text above. Such a sequence >>of numbers are not (pseudo) random numbers anymore, because it would not show >>the expected statistics for this properity! >> > >Randomness is not important. I didn't want to imply anything different. I just mentioned it, because people seem to call this still a sequence of reandom numbers. >[...] I talked to my algebraic coding theory >professor about this, [...] Perhaps you can have another talk with him :-) Generating hash keys reminds me very much of lagged fibonacci RNGs, where you get a new random number, by doing some operation on previously generated numbers. Often something like r = state[n-k] op state[n-l]. For this class of RNGs, it is well known, that using xor as op yields much worse results, than using for example subtraction (this is the RNG used in Crafty), or addition. Overflow and underflows would wrap around, just like the typical unsigned artithmetics. So, would it be preferable for hashkeys to use addition as well (and subtraction to clear a piece from the board). It should use almost the same time as the xor method? Regards, Dieter
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.