What's wrong with this EPD?

[Jouni]

Chessbase says illegal, but Arena accepts:

[D]Nb1QN3/1pp1PPp1/1rnn1p1P/1Pkr2B1/P1pRpK1p/2P1P2B/2qb2Pp/4R3 w - -

Jouni

[sje]

The full six field FEN is:
[D]Nb1QN3/1pp1PPp1/1rnn1p1P/1Pkr2B1/P1pRpK1p/2P1P2B/2qb2Pp/4R3 w - - 0 1

It is legal FEN.

[George Tsavdaris]

The full six field FEN is:
[D]Nb1QN3/1pp1PPp1/1rnn1p1P/1Pkr2B1/P1pRpK1p/2P1P2B/2qb2Pp/4R3 w - - 0 1

It is legal FEN.

Actually that would be illegal too.
How can you reach this position in 1 move?

Replacing w - - 0 1 with w - - 0 100 for example i guess would make it legal(if there is a sequence of moves that reaches this position from the initial).

[Dann Corbit]

White has all but one of his chessmen on the board.
How did the black pawns move to those locations.

As a retrograde position, it is not possible. It's pretty clever that an interface caught it.

[hgm]

'Annoying' would describe it better than 'clever'! If it really refuses it for this reason, it is just needlessly restricting its usefulness to the user. Nothing worse than pedantic software...

[sje]

FEN describes all chess positions including those that cannot be reached from the initial array.

[Evert]

White has all but one of his chessmen on the board.
How did the black pawns move to those locations.

Same question about the white pawns on the 7th rank, actually.

[Jouni]

Stockfish 2 isn't doing anything in this position, a bug?

2011-03-20 20:04:12.666-->1:isready
2011-03-20 20:04:12.666<--1:readyok
2011-03-20 20:04:12.791-->1:position fen Nb1QN3/1pp1PPp1/1rnn1p1P/1Pkr2B1/P1pRpK1p/2P1P2B/2qb2Pp/4R3 w - - 0 1
2011-03-20 20:04:12.791-->1:go infinite
2011-03-20 20:04:12.807**Time for starting analysis:640
2011-03-20 20:04:14.304-->1:stop

Jouni

[sje]

Stockfish 2 isn't doing anything in this position, a bug?

Maybe not, depending on interpretation. Black's bishops are running on the same color squares but neither is the result of a promotion.

Then again, if a program thinks that a position is broken, then the program should note that fact in its output or at least refuse to accept the broken position.

Another subtle problem that can hang an unsuspecting program: triple check. This can't happen in any position from the initial array, but it can happen in a setup position. It's possible for a program's check evasion routine to say "if my king is attacked but if the number of attacks is not two then I assume the number is one". That would be okay for normal play, but not for a maliciously specified busted position.

[Sven]

Stockfish 2 isn't doing anything in this position, a bug?

Maybe not, depending on interpretation. Black's bishops are running on the same color squares but neither is the result of a promotion.

Then again, if a program thinks that a position is broken, then the program should note that fact in its output or at least refuse to accept the broken position.

Another subtle problem that can hang an unsuspecting program: triple check. This can't happen in any position from the initial array, but it can happen in a setup position. It's possible for a program's check evasion routine to say "if my king is attacked but if the number of attacks is not two then I assume the number is one". That would be okay for normal play, but not for a maliciously specified busted position.

I think it is acceptable if an engine rejects an FEN of a position that cannot be reached by any sequence of legal move from the initial position (possibly including all 960 initial positions of Chess 960 if you like). This implies that situations like "triple check", "two like bishops where none of them results from promotion", or "pawn structure would have required more captures than enemy pieces are missing" can be rejected.

I agree that the way how such a position is rejected should be like you mentioned, i.e. not crashing but either sending an appropriate output or refusing to play or analyze from that position.

An engine can make assumptions like the one you mentioned about the number of checks if it is strict enough in rejecting "illegal FENs". Here the term "illegal FEN" is not precise, of course, since it is not the FEN itself which is illegal (the PGN standard describes the format only) but the position described by the FEN is unreachable based on FIDE chess rules.

It is a separate topic whether engines should also support analysis (or even game playing) for such unreachable positions. I would not be surprised, however, if most of today's competetive engines stick to not supporting these since playing strength in standard games often has the main focus.

Sven