Illegal move : who is guilty ?

[Olivier Deville]

In a game between JanWillem and Plywood played last night, JanWillem as white plays 29.Rf4-f5. There are 2 white Rooks on the f file, but only the rook at f4 can go to f5 (the other rook is on f1).

By looking at the debug file, it seems that JanWillem sent R4f5. But I am not expert in debug files because of my Fritz background so I need some advice.

Here is the relevant part of the arena debug.

Olivier

Code: Select all

1726734<1:move f4f5
1726734*1*Coup trouv?:Tf4-f5
1726812<1:NPS: 1070462
1726937*2*Start calc, move no: 57
1727234>2:time 33341
1727281>2:otim 39268
1727281>2:usermove R4f5
1727297<2:Illegal move: R4f5
1727328<2:3 72 3 1628 Rh6 Kf2 Ne5
1727344<2:4 79 3 5227 h5 Re1 Nc5 Kf2
1727359<2:5 78 6 16223 h5 Re1 Ne5 Rxf6 Rxf6
1727422<2:6 78 14 52654 h5 Re1 Ne5 Rxf6 Rxf6 Rxe4
1729016<2:7 54 171 851697 Kd8 h5 Nxh5 Rxf7 Rg7 R7f5 Ng3
1731734<2:8 60 443 2136072 Kd8 Be2 Rh6 Bc4 Ke8 Kf2 Ne5 Bb3
1740453<2:9 61 1315 5893444 Kd8 Rb1 b6 axb6
1753437<2:move Kd8


[Uri Blass]

Movei knew in the past only to read f4f5 and even now after it knows to read pgn when it runs under winboard it expects f4f5 and not R4f5.

I see no reason that engines will not use full notation like f4f5 in engine-engine games.

You only ask for trouble by not doing it.

Uri

[Thomas McBurney]

In a game between JanWillem and Plywood played last night, JanWillem as white plays 29.Rf4-f5. There are 2 white Rooks on the f file, but only the rook at f4 can go to f5 (the other rook is on f1).

By looking at the debug file, it seems that JanWillem sent R4f5. But I am not expert in debug files because of my Fritz background so I need some advice.

Here is the relevant part of the arena debug.

Olivier

Code: Select all

1726734<1:move f4f5
1726734*1*Coup trouv?:Tf4-f5
1726812<1:NPS: 1070462
1726937*2*Start calc, move no: 57
1727234>2:time 33341
1727281>2:otim 39268
1727281>2:usermove R4f5
1727297<2:Illegal move: R4f5
1727328<2:3 72 3 1628 Rh6 Kf2 Ne5
1727344<2:4 79 3 5227 h5 Re1 Nc5 Kf2
1727359<2:5 78 6 16223 h5 Re1 Ne5 Rxf6 Rxf6
1727422<2:6 78 14 52654 h5 Re1 Ne5 Rxf6 Rxf6 Rxe4
1729016<2:7 54 171 851697 Kd8 h5 Nxh5 Rxf7 Rg7 R7f5 Ng3
1731734<2:8 60 443 2136072 Kd8 Be2 Rh6 Bc4 Ke8 Kf2 Ne5 Bb3
1740453<2:9 61 1315 5893444 Kd8 Rb1 b6 axb6
1753437<2:move Kd8


Hi Olivier!

This is an Arena bug! It has been known for quite some time that Arena SAN output does not conform to SAN rules. I have reported this bug to Arena almost a year ago but I suspect this is low priority to them. I would report this bug to Arena myself again but they won't accept planet Earth as my address.

You can see in the log that Arena received the move "f4f5", then Arena interpreted and converted this to "R4f5" (SAN) for the other engine. The other engine probably has a correct implementation of SAN and therefore could not find this move. The correct SAN for this is "Rf5".

I hope the Arena developers will find time to fix this problem because I don't believe that programmers should be forced to implement an incorrect version of SAN just so that it will work with Arena.

Winboard has SAN implemented correctly.

Cheers,
Tom.

[Olivier Deville]

Hi Tom

Thanks a lot for your explanation

I will report the bug to the Arena team, and as for this game, I think it should be replayed.

Best

Olivier

[Thomas McBurney]

Hi Tom

Thanks a lot for your explanation

I will report the bug to the Arena team, and as for this game, I think it should be replayed.

Best

Olivier

Thanks Olivier,

The specifications for SAN can be found at - http://www.chessclub.com/help/PGN-spec

You will find the relevant specs at the section called '8.2.3.4: Disambiguation'.

Cheers,
Tom.

[Uri Blass]

In a game between JanWillem and Plywood played last night, JanWillem as white plays 29.Rf4-f5. There are 2 white Rooks on the f file, but only the rook at f4 can go to f5 (the other rook is on f1).

By looking at the debug file, it seems that JanWillem sent R4f5. But I am not expert in debug files because of my Fritz background so I need some advice.

Here is the relevant part of the arena debug.

Olivier

Code: Select all

1726734<1:move f4f5
1726734*1*Coup trouv?:Tf4-f5
1726812<1:NPS: 1070462
1726937*2*Start calc, move no: 57
1727234>2:time 33341
1727281>2:otim 39268
1727281>2:usermove R4f5
1727297<2:Illegal move: R4f5
1727328<2:3 72 3 1628 Rh6 Kf2 Ne5
1727344<2:4 79 3 5227 h5 Re1 Nc5 Kf2
1727359<2:5 78 6 16223 h5 Re1 Ne5 Rxf6 Rxf6
1727422<2:6 78 14 52654 h5 Re1 Ne5 Rxf6 Rxf6 Rxe4
1729016<2:7 54 171 851697 Kd8 h5 Nxh5 Rxf7 Rg7 R7f5 Ng3
1731734<2:8 60 443 2136072 Kd8 Be2 Rh6 Bc4 Ke8 Kf2 Ne5 Bb3
1740453<2:9 61 1315 5893444 Kd8 Rb1 b6 axb6
1753437<2:move Kd8


Hi Olivier!

This is an Arena bug! It has been known for quite some time that Arena SAN output does not conform to SAN rules. I have reported this bug to Arena almost a year ago but I suspect this is low priority to them. I would report this bug to Arena myself again but they won't accept planet Earth as my address.

You can see in the log that Arena received the move "f4f5", then Arena interpreted and converted this to "R4f5" (SAN) for the other engine. The other engine probably has a correct implementation of SAN and therefore could not find this move. The correct SAN for this is "Rf5".

I do not understand why arena needs to convert f4f5 to a move that is harder for engines to understand.


I guess that every engine that understands SAN understands f4f5 because it is clearly easier task to understand f4f5 than to understand SAN.
Movei works under arena inspite of not supporting SAN so I guess that the problem is not with arena but that arena was told that the engine needs to get SAN.

Can you give me a single reason that an engine will accept only SAN notation and not the more ugly f4f5 notation?

Uri

[Laurens Winkelhagen]

Hi Olivier,

I can confirm that this must be an arena bug, because thusfar I have not attempted to Implement SAN output;-) My engine only produced full-notation output at this time, so f4f5 instead of R4f5.

Thanx again for running your tournament. I would like to take this opportunity to notify you of an error in JanWillem 1.09, more specifically an error in the openingbook causing JanWillem to send an illegal move message. Because I can't predict when I will be able to compile JanWillem 1.10, I must ask you to manually modify to openingbook. If you would, could you please open the book in wordpad, find the line:

Code: Select all

r1bqkbnr/pp1ppppp/2n5/2p5/4P3/5N2/PPPP1PPP/RNBQKB1R w KQkq -

and substitute in the line below: c2c3 for b2c3?

Anyway, I have a feeling that my openingbook is bringing JW down in playing strenght (currently testing that). For the strongest play of JanWillem it might be best to delete the whole openingbook altogether, or perhaps replace it with the much smaller book that came with JW 1.07;-)

I hope I will be able to produce a stronger book soon,

Kind regards, Laurens.

[Laurens Winkelhagen]

btw, I'm curious to know how you will resolve this issue? Replay the game?

Greetings, Laurens.

[Olivier Deville]

Yes Laurens, the game is being replayed right now and it is even broadcasted ! You can come see it if you are quick

IP 83.157.47.162, port 16043

Olivier

[Laurens Winkelhagen]

I Really don't mean to use this forum to chat, but I must say... unfortunately I can't seem to connect:-( and because I connected once I was hoping that there might be a small mistake in typing the IP address?? Otherwise I'm at a loss again

anyway, I'll see the result on your site soon:-)

Laurens.

[Olivier Deville]

I apologize Laurens, I gave you the old IP

The good one is 83.157.39.121 port 16043

Hurry up, the game is almost finished

Olivier

[Uri Blass]

Hi Laurens,

If you even did not try to implement san notation then I guess that Olivier did something wrong(otherwise I could read problems with movei under Arena because I also did the same as you and I never readed about similiar program with movei).