Author: Heiner Marxen
Date: 12:49:59 05/19/99
Go up one level in this thread
On May 19, 1999 at 15:15:31, KarinsDad wrote: >On May 19, 1999 at 11:32:51, Heiner Marxen wrote: > >>On May 18, 1999 at 08:02:09, Peter Klausler wrote: >> >>>On May 17, 1999 at 21:15:59, Dann Corbit wrote: >>> >>>>You would also need to store e.p. square, if any, I think. Ten bits more >>>>[5ep+4castle+1stm]gives 174 bits. >>> >>>Two bits are all you need for the en passant file, actually. >>>There can be at most four pawns of the opponent on his >>>fourth rank next to a pawn of the player (on his fifth). >> >>Wrong. There can be 5 such pawns. E.g. 8/8/8/8/PpPPpPPp/8/8/8 b - >>Even when we know already that there is an e.p. possible, we still >>have to distinguish up to at least 5 cases. > >No, you only need 1 ep bit (based on the structure that this original statement >was made). Sure. Ok. I did not question the counting of bits, but wanted to correct Peter's "There can be at most four pawns...". >Because, if ep exists, then you can add 4 more bits to the structure >to represent the square involved (3 bits) and the direction of capture (1 bit, >a file to b file, or b file to a file type of thing) and you can remove 6 bits >from the original structure for the pawn that can be taken by ep and the pawn >that can take (if you have 2 pawns that can take, you do not care since the >other one still shows up in the original structure). Hence, an en-passant case >takes up 2 bits less (+4 -6) than a non-en-passant case. > >So, only 1 ep bit is needed. That is correct IMHO. And I like these tricky encodings :-) >KarinsDad :)
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.