Author: Dann Corbit
Date: 23:56:01 01/30/02
Go up one level in this thread
>On January 30, 2002 at 21:11:25, Angrim wrote: >[snip] >>Hmm, this made me stop and think a bit. If you could actually skip >>over the positions that are covered by your rule, while still using >>the current index based methods, the resulting savings would indeed be >>nice. However, I do not see any way to do this. To calculate a >>positions index, you need to know how many positions that would >>otherwise have been in this file would have been before it in the >>file. For the "kings not touching" rule, this can be calculated >>directly based on the positions of the two kings. If there is a >>way to calculate this for the sliding piece rule, I have overlooked >>it. >>If you do come up with a practical way to do this, it could make >>this whole thread worthwhile :) > >The permutations are a simple algorithm. I suspect that a bit of math may make >such a calculation possible. > >It is possible that the idea will fizzle. But I would like smart people (like >yourself) to give it a bit of thought and perhaps something useful will fall out >of it. > >Forget about the encoding scheme of exactly how a position is stored. It was >just Les' way of demonstrating his basic idea. The keen idea is the reduction >that comes from looking for a solution and recognition that many different roads >all lead to Rome. > >As a sidelight -- the postions I posted earlier for KQKnn turn out to be >uniquely correct (all the generated mates are optimal : none of the Nalimov >lookups for the generated mates are shorter than the extrapolated ones by the >simple algorithm). It is conceivable that some tablebase files might be optimal >or nearly so. A mapping to the subset is very simple, when you think about it. Just form the complete list of key positions. Then form an n-bit perfect hash for these positions, where n is large enough to hold all the entries. Then, during a game, you just form the key then form the hash. Simple
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.