Crafty V11.3
Robert Hyatt
I have just copied Crafty version 11.3 to the ftp machine
ftp.cis.uab.edu/pub/hyatt/ These files did *not* replace
crafty.zip and the other jakarta release files.
The new files are all named crafty-11.3.* so that 10.18 will
remain on the ftp machine as a known stable release.
11.3 has the new (not finally tweaked however) fractional ply
extension code, modified king safety code, and a fix so it will
annotate a game sent by ICC (ICC was emailing PGN games with
the moves like this: 1.e4 e5 2.Nf3 etc..., with no space
between the . and the move.) Crafty will now read these fine
and will also read those with spaces...
other minor fixes include full pgn header support so that
annotated files have the original header stuff in them so
you can tell what's what...
Walter Ravenek
In article <54isir$j...@juniper.cis.uab.edu>, hy...@crafty.cis.uab.edu
(Robert Hyatt) wrote:
> 11.3 has the new (not finally tweaked however) fractional ply
> extension code ...
Bob,
It seems like some ideas are in the air...
Last weekend I also changed to fractional ply extensions, so I had
a look at your code in Crafty to see how you do it. It was only a
short one, but I got the impression that you don't look for the
maximum extension depth if more than two extension criteria are met.
Am I right?
Walter Ravenek
--
Walter Ravenek
Robert Hyatt
In article <ravenek-ya0231800...@news.vu.nl> you wrote:
: In article <54isir$j...@juniper.cis.uab.edu>, hy...@crafty.cis.uab.edu
:
Yes. I assume (sometimes incorrectly no doubt) that extensions will
wash out as the tree deepens. Currently I am extending 1 ply for out
of check, 1 ply for recaptures, 3/4 ply for one legal response to check
and 3/4 ply for passed pawn pushes. I have several more in Cray Blitz
that I'm not yet using, but there are few real limits here other than I
don't let *all* of these kick in at the same time for obvious reasons.
I do want the search to terminate *eventually*... :)
Bob
Robert Hyatt
I've just put new source and executables for 11.3 on the ftp machine
ftp.cis.uab.edu/pub/hyatt/ This fixes a problem that only affects those
running crafty on a chess server. We've seen Crafty's aggressive use of
time cause problems when some humans try to lock things up and then "sit"
and move very quickly trying to run crafty out of time. This fix will
make Crafty speed up as the opponent begins to pull ahead on time, which
effectively avoids flagging now. I've watched one game against a player
named "insight" where he tried this again after winning a couple by flagging,
and he actually accused me of "interfering" with the game. :) I told him
what I'd done, he tried it one more time, then went back to playing "chess."
As I've said before, playing "automatically" on a server is a huge challenge,
from a chess-skill point of view *and* from many other angles too, such as
automatically allocating time for a search, etc. Time allocation is a
sophisticated problem requiring sophisticated algorithms to avoid problems.
Bob