Author: Dann Corbit
Date: 11:55:16 01/30/02
Go up one level in this thread
On January 30, 2002 at 05:07:02, Uri Blass wrote: [snip] >and side to move that make 16 "reflections" based on the definition of having >practically the same position(this is not the definition of Les). > >b1 a2 a7 b8 g1 h2 h7 g8 are symmetric when there are no pawns and in all of >these cases you can change the side to move. In the encoding scheme that Les invented, you also *multiply* this by the number of ways the sliding piece can go to the solution square. Suppose that a rook can mate in 12 if he slides to the solution position. There are up to 13 other squares he can move to the same solution sqare from other than the square that he is sitting on. So in this case, there are up to 16 * 13 = 208 board positions that will all have a mate in (12 or better). Now, Les only saves *one* of these positions [the smallest one lexically]. All of the others can be generated from that one. Along with all of their key moves and all of their centipawn evaluations. With a queen, it is even more drastic. So that is how he saves space. Work it out for KQK and see what the savings is. A Nalimov tablebase file gives superior information. But the format Les makes is so much more compact, it might be possible to keep a 5 man set completely in memory. (That's a wild guess, I have not done the math -- but suppose that it could compress down to one gigabyte -- you can get 4 gigs of ram for $600).
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.