Did i found a new spell chess bug ??? Oops...

Chess.com variant server is sometimes a little bugged. In spell chess we already found a bug which is explained on the page linked. But now, i think i have found a new bug (this one won’t have any importance toh, or just for people trying to get nice checkmates). Please consider the very simple following line :

1. e4 e6

2. freeze@c7Bb5 freeze@c5Bb4

3. c3 c6

4. cxb4 cxb5

5. Nc3 Nf6

6. Nxb5 a6

7. freeze@e7Nd6 freeze@e5Nc6

8. Qh5+ Ke7

9. jump@d6Qc5+

If you put this line in spellchess.win, it’s gonna add a little # (it’ll be considered as a checkmate), while if you write this line in the variant server, it’s not gonna add the # because it thinks jump@d7Qxd6 removes the piece we jump over and isn’t check anymore.

To conclude it seems like chess.com considers that the jump applies on a piece and not on a square.

Tell me in the forum what you think of this new bug/rule précision, and new articles soon !

@Fleex255

@Delhy

Nice find! Quick link to the position on spellchess.win: https://spellchess.win/#!moves=DBw0LPIFIeI9GQoSMioSGSohARI+LRIhMCj0ISvkOSoDJzw0qyci

I'm not aware of an explicit ruleset from chess.com that says whether jumps apply to pieces or squares. I always thought it applied to squares, so if it's actually supposed to apply to pieces, that's a bug in spellchess.win.

However, chess.com is inconsistent about this. Consider this silly line where White applies a jump to a black pawn and takes it:

1. e4 e6
2. Bc4 d6
3. jump@e6&Bxe6 Bg4

After White's move 3, both sites show White's bishop as being jumped even though the jump was played onto Black's pawn. Black is allowed to move through it on their turn, so the jump stayed on the square. Also:

• In your line, if Black wasted a jump on move 8, chess.com does consider 9. jump@d6&Qc5 to be mate, so jumps definitely expire after testing for check.
• Wasting a jump on the piece you're moving leaves the empty square jumped. The piece is not jumped in its new position.

At some point, I might look for a programming mistake in the variants client to see if it's a bug or just a super-specific rule. It really seems to me that jumps apply to squares, though.

Fortunately, I don't think this can ever matter tactically, since the crushing move (9. Qc5 in this case) without the jump is spell-mate (unstoppable king capture next turn).

About the silly line I suppose this works ‘cause the piece hasn’t been replaced by another one, in my line you really remove the jumped piece…

For the waste of move 8 for black could you precise a line ? If black doesn’t play Ke7 and white has time for jump@d6Qc5 black can sac the queen with jump so it’s not checkmate. Do you mean if the queen was a knight it would be checkmate ?

For the 3rd point you’re right with an exemple like 1. jump@e2 e4.

For the matter, I think it’s pretty interesting to see what chess.com considers like checkmate or no.

According to me spellchess.win is way less bugged than chess.com and it’s doing the right analysis of checkmate for me.

The conclusion would more be : if you are in check through a piece because of a jump in it, you can ignore the check and remove the jumped piece.

Tell me what you think of this !

Consider the silliest line so far :

1. e4 d6

2. Nf3 f6

3. Ne5 g6

4. Nc6 a5

5. jump@c6Bb5+ the move bxc6 is allowed

My idea seems right no ?

About "jumps definitely expire after testing for check", I meant Black playing Ke7 with a pointless jump.

Thanks for finding that bxc6 case, that's a really good example. "Capturing" the jumped piece apparently also applies to an empty square, like in this nonsense line where the queen jumps herself and steps back, leaving an empty jumped square. Chess.com allows a piece to interpose on the jumped square, but spellchess.win considers it mate because the square is still jumped at the end of Black's turn.

For an authoritative source, I just found the variants update announcement from April 2025, which mentions this as a bug fix: "Captures of pieces under jump spell are now allowed if spell ends the next move." With the empty square test case, I would actually express the rule as: "Moving onto a square jumped by your opponent immediately cancels the jump."

So chess.com is working as intended, and spellchess.win is missing a rule. I will try to fix this by the next monthly data update.

Thanks again for finding this!

That’s very intersting, I did not know about this rule, as unuseful as it is. By the way abiut “jumps definitly expire after testing for check”, I think your engine doesn’t see those weird mates where you use a jump just to checkmate, like in your line already mentionned, the engine sees mate in two instead of mate in one with the last move.

Right, my engine never considers moves that spend a jump without moving through the jumped square. Those moves are always useless unless they're mate in one (or maybe in an incredibly contrived stalemate scenario). Skipping them lets the engine spend its time evaluating useful positions, which is very important for strength. Reporting a weird M1 as M2 is a tradeoff I'm OK with for now.

Yes I agree those mates aren’t important and looking for them would make the engine lose depth and time.

This is fixed on spellchess.win. I added a sentence to the "rules reference" section so we have all the rules in one place. Thanks again for the report.

Great !