Re: [Scid-users] Crafty w/ SCID and Behaviour of header search

[Shane H. <sh...@co...>]

On Sun, 14 Jul 2002, Denis J. Navas Vega wrote:

> I wish to inform the following SCID behavior with regard a search using headers
> and difficulties using crafty to analyze positions:
> 
> 
> Search by headers:
> -------------------
> 
> I recently did a search by headers, using the game information from puzzle #2 of
> London Chess Center boletin  Vol.1, #2 - July 12, 2002
> 
> The puzzle #2 has the following data:
> 
> 2b2rk1/pr3n1p/3p4/3Np2n/1PP2P1q/8/P4PBP/1R1Q1RK1 w - - 0 21
> Romanishin - Tibensky, Kaskady, 2002.
> with the solution:  21.Qxh5! 1-0 21...Qxh5 22.Nxf6+ Kh8 23.Nxh5
> 
> Besides player's last names, if I set the year as "2002" I get no game at all,
> with "2002." (see the point at the end of  the year) I retrieve the mentioned
> game.
> 
> Is this the right behaviour?  Or the software expects to add the month?

Well, it depends what the PGN representation looks like. It sounds like
you are storing the site and year in a player (White or Black) field.
I'm not sure why adding/removing the . after the 2002 should make a
difference though.
 
> Using Crafty:
> -------------
> 
> A)  Getting an evaluation with changing sign
> 
> I recently installed Crafty 18.15;  after that install I get a different
> behaviour from the previous version.  With the new installation, evaluation
> changes from positive/negative acording to the side on move in SCID analyze
> window.  But in Crafty window the evaluation sign preserves it signs regardless
> the side
> on move.  Previous to 18.15 version I always obtained the same sign with the
> position evaluation, regardless of side to move, wich is the expected behaviour
> acording to Crafty manual.

Every Winboard engine except crafty (as far as I know) scores from
side-to-move perspective, but crafty uses white perspective and Scid tries
to do that. I just tried Scid with Crafty 18.15 and it worked fine.
I can't comment on possible changes in Crafty, but I can point out
that Scid looks for a line from the engine starting with "Crafty vXX.YY"
where XX.YY is some version number, and if that number is >= 18.0,
it assumes the engine prints scores from the White perspective.
Perhaps you could check the engine1.log file, and see if near the
start you get something like this:

Engine: Crafty v18.15
NOTE  : Seen "Crafty"; assuming analyze and setboard commands.
NOTE  : Crafty version is >= 18.0; assuming scores are from White perspective.

If you don't, it means Scid will be assuming the engine is using the
perspective of the side to move for evaluation scores.
 
> B)  Time per ply reported, as 0.00 always
> 
> I ignore if Crafty have changed the way it works, but I suspect it did, because
> I had to make a slight change to crafty.rc in order to get it inform correctly
> the time spent in each play.  I changed the line "time cpu" in crafty.rc to
> "xtime cpu" (instruction invalid) to get the the accumulated time per ply
> correctly informed, instead of obtaining 0.00 sec in every ply.  Obviously this
> is a change introduced by crafty author.

I have no idea, but I usually start crafty in Scid with no rc file and I
do not know what the "time cpu" command does.

> C)  Numbering of moves in analyze window with Crafty
> 
> The second problem I am having with Crafty & Scid, is that I am not getting the
> moves number if I set a position using "Edit" + "Set up start board" taking care
> to fill fields of move number, castling rights and en pasant file.  I get an
> analyze identical of what comes from the start position (eg.  1.e4 e6 2.d4 d5
> ...).   If I analyze a position from a loaded game, the move numbering is ok in
> the analyze window.

That looks like a bug in Scid; it sends a "mn" command to Crafty to set
the move number (mn is not a Winboard standard command, it is one only
crafty understands, so it won't be sent unless Scid thinks the engine
is a version of Crafty) but Scid seems to be sending the wrong number.
Hopefully I can fix that sometime.

Cheers,
Shane