Author: Gerd Isenberg
Date: 04:29:06 12/20/02
Go up one level in this thread
On December 19, 2002 at 19:36:18, Andreas Herrmann wrote: >On December 19, 2002 at 18:48:45, Gerd Isenberg wrote: > >>On December 19, 2002 at 16:12:24, Andreas Herrmann wrote: >> >>>>Hi Andreas, >>>> >>>>i think Rcf2 is not incorrect here, don't considering invalid moves due to >>>>pinned pieces. Better to have redundant rather than missing information. Your >>>>SAN-parser should handle both Rcf2 and Rf2 correctly (but not Rff2 ;-) >>>>I guess writing the move is more or less implementation depending, even if an >>>>exact definition about this issue exists. >>>> >>>>Regards, >>>>Gerd >>> >>>Hi Gerd, >>> >>>yes you are right. Holmes also accepts moves without the check "+" or "checkmate >>>"#" characters or with the remark characters "!" or "?". >>> >>>But in the above case i have problems, because in the root i generate only legal >>>moves (no pseudolegal ones). And from all possible legal moves i generates a >>>SAN, a LAN and coordinate notation list to compare them with the input move. If >>>the input is not a legal move (SAN, LAN or coordinate input) than i try to >>>delete or add the above characters (+,#,!,?) and compare again, that's no >>>problem. >>> >>>To produce a notation like Rcf2 i have to generate also pseudolegal moves in the >>>root, because only in this case i have 2 possible rook moves to the same target >>>square. I think this extra work stands in no relation to this rare case. >> >>Yes i see. But string compare? >>If you parse a move string from left to right, character for character, i would >>suggest a kind of finite state machine, where a current (initial) state and the >>next character define some action (eg. tagging or shrinking the move list) and >>an next state. > >Yes Dieter Bürßner explaned me something similar in the Winboard forum, but i >think normaly a GUI has to send and export the correct SAN notation. >Perhaps i will change my code after the IPCCC. But at the moment i have to find >500 or more ELO in Holmes to beat Isichess in Paderborn ;-) > A bit overstated ;-) >> >>Nice to have bitboards ;-) > >At the moment i'm just searching for a faster solution for calculation the SEE >value. But i think this would be easier and faster with bitboards. >Can you give me some hints? Don't use explicite SEE. I have some bitboards with defending properties, like defended by pawn, by light pieces, defended more than once, or not defended at all, etc. This is also usefull for none captures. I combine these with MVV/LVA ordering and only prune obviously loosing none checking captures like Queen takes Pawn defended by Pawn in qseach. Gerd > >Currenty i have changed my move ordering from SEE back to MVV/LVA because my SEE >code was much too slow. > >> >>> >>>have a nice day >>>Andreas >> >>and a nice night, >>Gerd >> >> >>> >>>PS: Have you heard something from Jochen, because he stands not in the IPCCC >>>participants list? So i'm fighting only with Matador for the "Rote Laterne" next >>>year :-) >> >>No, heard nothing about Jochen. >>But the number is still odd, so hopefully... >>A damned strong field. > >Yes perhaps stronger than the last WM. > >Andreas
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.