Computer chess analysis • page 5/6 • Game analysis • lichess.org

"I wouldn't play that move for 10 bucks!"
Meh I prefer the +- or 2.01 myself dont you?

Additionally I think both thibs and someone have excellent ideas.
If I recall gameknot.com had a queue so that things arent over complicated for their analyze feature.

"I wouldn't play that move for 10 bucks!" Meh I prefer the +- or 2.01 myself dont you? Additionally I think both thibs and someone have excellent ideas. If I recall gameknot.com had a queue so that things arent over complicated for their analyze feature.

thibault

#43

exact. I'd add that the chess server has 8 cores, so I can process 6 games in parallel (and save 2 cores for the AI games).

Also I'm making it distributed, so if someone wants to install a processing server, he will be able to take some of the load. That will be fun.

exact. I'd add that the chess server has 8 cores, so I can process 6 games in parallel (and save 2 cores for the AI games). Also I'm making it distributed, so if someone wants to install a processing server, he will be able to take some of the load. That will be fun.

ojb

#44

sure thing, having looked at it again it is pretty nasty stuff, but there are some pretty clear pathways to improve it I think, currently I have to preprocess the pgn into a list of algebraic moves - better would be to have it use a FEN representation, then things open up in terms of caching positions you evaluated in a different session...

I wrote it around a year ago and it has been handy on my game-posting site (chesspo.st). sadly that site is currently fairly broken and due to getting a real job I haven't had time to play with fun sideprojects, much less refactor this great clanking beast :-) but maybe it will prove of some use.

prerequisites are expect, pgn-extract (i have v16.7 here) and a uci engine.

"./analyse my.pgn" should give you my.pgn.out (after LOTS of spam)
uci-analyse is the expect script that does the magic, enter at your peril ;)

bonne chance!

http://ojb500.co.uk/analyse
http://ojb500.co.uk/uci-analyse

sure thing, having looked at it again it is pretty nasty stuff, but there are some pretty clear pathways to improve it I think, currently I have to preprocess the pgn into a list of algebraic moves - better would be to have it use a FEN representation, then things open up in terms of caching positions you evaluated in a different session... I wrote it around a year ago and it has been handy on my game-posting site (chesspo.st). sadly that site is currently fairly broken and due to getting a real job I haven't had time to play with fun sideprojects, much less refactor this great clanking beast :-) but maybe it will prove of some use. prerequisites are expect, pgn-extract (i have v16.7 here) and a uci engine. "./analyse my.pgn" should give you my.pgn.out (after LOTS of spam) uci-analyse is the expect script that does the magic, enter at your peril ;) bonne chance! http://ojb500.co.uk/analyse http://ojb500.co.uk/uci-analyse

ojb

#45

@legend the $10 stuff are PGN NAGs (Numeric Annotation Glyphs). when viewed in a pgn viewer they turn into the typical +/- evaluation symbols

thibault

#46

holy guacamole, that looks crazy. I never heard about this "expect" language before. Let's give it a try!

Mephostophilis

#47

DIS HEAH'S A TEN-DOLLAH MOV

f3 :O

$0.99

?XD

DIS HEAH'S A TEN-DOLLAH MOV f3 :O $0.99 ?XD

Mephostophilis

#48

But yeah the static "+-" layout, or estimating in centipawns/pawns should work great. I take it the '$' symbol is supposed to be a variable/register access token to the virtual machine to retrieve the requested value to subset into the string, otherwise it's a piece of crap. I guess where C/++ uses "%" to do this, Expect uses "$".

But I never heard about Expect before. The best language for giving the results I expect is assembly imo. :P Effing compilers.

But yeah the static "+-" layout, or estimating in centipawns/pawns should work great. I take it the '$' symbol is supposed to be a variable/register access token to the virtual machine to retrieve the requested value to subset into the string, otherwise it's a piece of crap. I guess where C/++ uses "%" to do this, Expect uses "$". But I never heard about Expect before. The best language for giving the results I *expect* is assembly imo. :P Effing compilers.

ojb

#49

Yeah, expect is crazy, i'd used it years ago to automate telneting (:D) into a remote site, but nothing quite this sick...

tsinnema

#50

Thibault, is that Crafty annotation stuff something that's built into Crafty or is it something external comparing the values of the engine moves to the values of the actual moves?

thibault

#51

what I shown is crafty raw output for the "annotate" and "annotateh" commands.

ojb: the more I dive into your expect script and the more I like it. This expect tool is completely amazeballs. I was trying to communicate with crafty and stockfish using conventional languages (namely scala and ruby) and found myself struggling with the asynchronous nature of the exchange. expect is made exactly for this purpose, that's cool.

I haven't managed to run it yet, going from an error message to another, tweaking your script. It's a lot of fun. Thanks again for sharing this script.

what I shown is crafty raw output for the "annotate" and "annotateh" commands. ojb: the more I dive into your expect script and the more I like it. This expect tool is completely amazeballs. I was trying to communicate with crafty and stockfish using conventional languages (namely scala and ruby) and found myself struggling with the asynchronous nature of the exchange. expect is made exactly for this purpose, that's cool. I haven't managed to run it yet, going from an error message to another, tweaking your script. It's a lot of fun. Thanks again for sharing this script.