Author: Dann Corbit
Date: 18:01:08 01/30/02
Go up one level in this thread
On January 30, 2002 at 20:54:25, Dan Andersson wrote: >>That would be 4 bits. Does that mean that the shrinkage is 2^4th = 16 fold? A >>simple way to accomplish that is to have 16 separate tables. Then you don't >>need any bits for castling rights. >The number of bits depend on how efficient the encoding is. Castling information >is implicit in the majority of cases. i.e. the king is off the castling square. >When calculating compression efficiency one takes into account the size of the >program. And that includes data also. For example: > You could have a compression that compresses Moby Dick to one bit. A one. And >all other cases a zero followed by other any other text (or text compressed by a >predetermined algorithm). But the size of that program would contain a >representation of Moby Dick. Thus not gaining any real benefit from it because >every other text will have an extra bit. > So the gain of separate coding of special cases comes into play only when >additional symmetries and recursive structures appear. > I do however think the idea of goal oriented sparse tables are good. As the >sliding piece idea from Lez. It is similar but simpler and more effective than >my idea of a sparse table containing enough position for search to find a match. >Even though they may need more than the optimal number of moves to win. One >complication remains, it is neccessary to ensure that the 50-move rule is not >breached. For this complication, simply do an ordinary tablebase lookup if the distance to mate + current non-reversible count is too large (IOW the projected mate is too far away). This idea of yours about sparse information search tables -- have you posted about it before? I don't remember reading anything about it. Could you give me the ten cent summary of how it works?
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.