Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Ten years later: revising CDS

Author: Steven Edwards

Date: 21:16:06 09/09/03

Go up one level in this thread


On September 09, 2003 at 23:51:32, Jon Dart wrote:
>On September 09, 2003 at 18:33:22, Steven Edwards wrote:

>Even in normal chess, if you take into account the previous history of the game,
>there are six possible castling statuses. Arasan names these as follows:
>
>                     CanCastleEitherSide,
>                     CanCastleKSide,
>                     CanCastleQSide,
>                     CastledKSide,
>                     CastledQSide,
>                     CantCastleEitherSide
>
>FEN can't currently distinguish these: it can only indicate whether or not
>castling is possible, not whether or not it has occurred.

Obviously, if a side has already castled, it cannot castle again.  So any
deficiency here is the lack of reporting that a castling has occurred in the
game history.  I think that this is more of a history issue than a position
status issue as it has no effect on any move generation or analysis.

Note that EPD has the "hv" (history variation) opcode that can be used to supply
the entire previous move history.  It's kind of like "sv" (supplied variation;
compare with "pv") but from the start of the game instead of from the current
position.

>>4. The centipawn evaluation operand type needs a mate score indication
>>correction.

The old "ce" has special scores to indicate mate-in-N and lose-in-N evaluations.
 That's not too bad, but unfortunately the special values used are broken and
represent an unwanted revealing of the internals of the analyzing program.
Example: "ce 32766" is a mate in 1 and "ce 32764" is a mate in 2 but "ce 32765"
is undefined and is, in fact, an error.  This can be fixed, and the SPCT has
such a fix already, but using the fix will likely break any software using the
old score interpretation.  That would be ungood.  So I think that it's time to
deprecate "ce" and let it fade into disuse by being supplanted with a correct
and new decimal pawn unit evaluation opcode and, probably, a new opcode for
representing special evaluations.  This would handle forced mates, forced
losses, and also identify forced draws by kind (e.g., stalemate, three time
occurrance).



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.