Author: Victor Zakharov
Date: 06:34:18 04/24/00
Go up one level in this thread
On April 24, 2000 at 08:06:58, blass uri wrote: >On April 24, 2000 at 08:00:45, blass uri wrote: > >>On April 24, 2000 at 06:57:50, Victor Zakharov wrote: >> >>>Theoretical calculations are very rough. Chess Assistant has algorithms of >>>variable length board compression (so called Direct tree). Results are >>>impressive. Average size is 12 bytes per position in Chess Assistant 4.0 and 11 >>>bytes in CA 5.0. And speed is good engough too. Two interesting observations: >>>there are a few positions that are packed in 56 bits (i.e. less than 64 bits and >>>less than one bit per a square!), and there are no positions packed more than 19 >>>bytes (144 bits). >> >>I believe that it is impossible because there are probably more than 2^144 legal >>positions. >> >>Uri > >I can add that my program worked hard to calculate an upper bound to the number >of legal positions and it only could prove that something like 2^156 is an upper >bound. > >I do not believe that chess assistant can compress every chess position with >less than 145 bits. > >There is probably a misunderstanding. > >Maybe it only compress part of the legal positions(for example legal positions >with no promoted pieces and 32 pieces) > >Uri Yes, I mean real chess positions. Sorry, I didn't read carefully start of discussion. 11 bytes - is average length for all positions in the database of 1 million games. Statistical laws were heavily used. Sure theoretically it is possible to create position that will be coded in 24 or 25 bytes (and may be more) in our program.
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.