Author: Johan de Koning
Date: 23:21:51 06/27/03
Go up one level in this thread
On June 27, 2003 at 18:02:44, Guido wrote: [clarification by JM] >Thanks for explanations. >It is very interesting the idea of generating TBs already compressed. >I suppose that the 20-25% of reduction is due to the chosen format that contains >more sequence of zeroes (from an "uncompressed" point of view), if compared with >Nalimov's TBs. These zeroes come back in the counterpart file (with reversed material) as non-zeroes, so effectively the data is split into a + and a - partition. The total amount of noise is still the same. I did however spend a lot of time researching the best way to compress typical EGDB data (and had a lot of fun doing it). >The main advantage of this format seems to me avoiding of 8-16 bits problem that >will arise with 6-men TBs, paying something with the reduction in information >contained in a single file. It is an advantage for *most* 6-man files. :-) Unfortunately some mates are longer than 508 ply, and they are brutally truncated by the current version of FEG. >I also suppose that the algorithm used by FEG is based on the retrograde move, >considering the relatively very short cpu time spent in generation. That's correct (besides a few other cool tricks to make it fast, a habit of chess programmers I guess). In theory backward generation addresses all the edges in the subgame graph at most once, while forward generation addresses them at least once. The difference increases with the number of passes. Regardless of the extra work, backward generation will win sooner or later. ... Johan
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.