Shartranj!

[Jim Ablett]

Right, I fixed Dabbaba so it won't freak out when the Shatranj fen starting position is loaded.
This a just a workaround i.e it wont accept any other Shatranj position other than the standard starting one..

http://www.mediafire.com/?dzmjjlet17g

Jim.

[hgm]

Ha! I have found a work-around after all. I can tell WinBoard to send a setboard to only one engine. After all:

The WinBoard is SMARTER!

The trick is to hide the WB protocol command in the "computer string", as this string can be specified for each engine individually. And, unlike the "init string", which is sent before the 'variant' command, and thus is not suitable to contain a 'setboard', the ComputerString is sent after all the other start-up commands of the game.

So what I do is install Pulsar with the command-line option

WBopt /%sComputerString="setboard rnbkqbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBKQBNR w 0 1\ncomputer\n"

in the engine manager. This then sneaks in a 'setboard' to the correct opening position just before the command 'computer', only for Pulsar. So I can play the tournament starting from the default opening position, sparing other engines (and in particular Dabbaba) from receiving a 'setboard' command.

That's right...? Thats's right!

[hgm]

Right, I fixed Dabbaba so it won't freak out when the Shatranj fen starting position is loaded.
This a just a workaround i.e it wont accept any other Shatranj position other than the standard starting one..

http://www.mediafire.com/?dzmjjlet17g

Jim.

OK, our messages crossed. But it is good to have a Dabbaba that understands 'setboard', as I might need to play from some shuffled openings to get enough independent games.

[Jim Ablett]

Hi Harm,

I just played a Shatranj test game, Fairy-max v Dabbaba where Dabbaba was left with a bare king being chased around the board.
Shouldn't either Winboard-F adjudicated a win or Fairy-Max claim one?

Jim.

[Jim Ablett]

Right, I fixed Dabbaba so it won't freak out when the Shatranj fen starting position is loaded.
This a just a workaround i.e it wont accept any other Shatranj position other than the standard starting one..

http://www.mediafire.com/?dzmjjlet17g

Jim.

OK, our messages crossed. But it is good to have a Dabbaba that understands 'setboard', as I might need to play from some shuffled openings to get enough independent games.

I'll try to get this sorted properly when I can.

Jim.

[adams161]

I'm not sure if this is because i'm using the 7-24-08 last modified
version of winboard_f, but I left pulsar on playing FR ( unrated ) and
when i got home i had two message boxes from winboard saying it couldn't
parse a move from ics. One of the message boxes was couldn't parse move
rg8. When i checked the debug log i noticed this is the game with
Iamhere, that that i realized pulsar had flagged in earlier today, but i wasn't
at home and couldnt look into it at the time tell i got home now. I did
have the moves mailed to me and one of the characteristics of the game
was move 1 was 1. 0-0 0-0, i.e. threechecksbot who had white and then
his opponent both castled on move 1. I will test more with the current
version you link to in hte prior email. The debug log is at www.adam16mr.org/stuff/winboard-pulsar-fr-8-14-08.zip

ps your mailbox said it was full when i tried to mail the log to you

Mike

[hgm]

I think this is a problem that has already been solved in the most recent version. The point was that the WB move disambiguation is only aware of the initial castling rights, not of the actual castling rights of the current position, and the move generator was not skeptical enough to test if an actual Rook and King were present on the board if the castling rights existed. So in the given position f1-g1 matched two moves: Rf1-g1 and O-O, ands the King started on f1 and the Rook on g1, and the internal representation of FRC castling is KxR. So f1-g1 was considered an ambiguous move.

George Tsavdaris pointed out this bug to me, and I fixed itt since (by making sure FRC castlings are generated only if K and R are present on from and to square).

[hgm]

I must say I am a bit disappointed in the level of play in this Shatranj touney. Let's face it, Fairy-Max (which would be paracticallty indistinguishable from micro-Max when playing normal Chess) is rather low-end engine on an absolute scale, even if it is a miracle for its size. And this Shatranj version is a lot weaker than the Chess version, since it is not even aware of the most important win condition (the rule that bare King loses), nor of the fact that being stalemated counts as a loss. In addition, Shatranj seems less tactical and more strategic than Chess, putting an engine with virtually no eval at a larger disadvantage. (From watching the games, it seems to me that a very important eval term would be the possession of Pawns that can promote on both colors, sort of a Shatranj equivalent to the Bishop pair.

The other engines are totally crushed by fairy-Max, however. Pulsar2007-9e-18 lost by 27.5-4.5, Dabbaba by 32-0. The match against the other Pulsar (using the opening book) is still going on, but not expected to give a significantly different result. Yet, Pulsar is not entirely bad, as it beat Dabbaba by 30-2.

Strange thing is that Dabbaba in the middle game, in my observation, very often seemed to have a winning advantage. It threw several games by sacrifycing its Rook to force stalemate, not aware that this counted as a loss. None of the engines is aware of the baring rule. In a dead drawn position (Like King+Rook vs King+Rook+Elephant) the side that is behind often offers a Rook trade, which would make it lose immediately because of baring his King. And equally often the opponent than does not take the trade! This is a really shameful display...

[Jim Ablett]

I must say I am a bit disappointed in the level of play in this Shatranj touney. Let's face it, Fairy-Max (which would be paracticallty indistinguishable from micro-Max when playing normal Chess) is rather low-end engine on an absolute scale, even if it is a miracle for its size. And this Shatranj version is a lot weaker than the Chess version, since it is not even aware of the most important win condition (the rule that bare King loses), nor of the fact that being stalemated counts as a loss. In addition, Shatranj seems less tactical and more strategic than Chess, putting an engine with virtually no eval at a larger disadvantage. (From watching the games, it seems to me that a very important eval term would be the possession of Pawns that can promote on both colors, sort of a Shatranj equivalent to the Bishop pair.

The other engines are totally crushed by fairy-Max, however. Pulsar2007-9e-18 lost by 27.5-4.5, Dabbaba by 32-0. The match against the other Pulsar (using the opening book) is still going on, but not expected to give a significantly different result. Yet, Pulsar is not entirely bad, as it beat Dabbaba by 30-2.

Strange thing is that Dabbaba in the middle game, in my observation, very often seemed to have a winning advantage. It threw several games by sacrifycing its Rook to force stalemate, not aware that this counted as a loss. None of the engines is aware of the baring rule. In a dead drawn position (Like King+Rook vs King+Rook+Elephant) the side that is behind often offers a Rook trade, which would make it lose immediately because of baring his King. And equally often the opponent than does not take the trade! This is a really shameful display...

Hi Harm,

Dabbaba definitely needs to understand some Shatranj fundamentals. Watching the test games has helped me
personally understand Shatranj play more and maybe I can pass that on to Dabbaba, so thank-you for running the test games.
I have sorted out the setboard command in Dabbaba, so now it will play from loaded positions.
http://www.mediafire.com/?0mnfhdhqgyx

Jim.

[adams161]

there is actually not that much specialized code in teh pulsar shatranj version. It also doesnt know about the bare king rule. shatranj was added becuase at one point it was a part of wild 29 ( the icc random wild were a wild is chosen at random ) and little to no effort was made to program for it other than i tried to strip out as much regular chess knowledge as i could. but i didnt really replace it with any new knowledge, so in a way its just a material count evaluate.

None the less in testing it plays rather well against the top icc shatranj players, but as i've been told in the past, shatranj is a wild that favors computers and is easily amendable to computer processing.