Please, let’s forget about bughouse, and illegal moves. I’d like to keep the discussion on-topic.
In that case, that e.p. flag (or “target square”, if you prefer) would need to be defined differently before it could be used to track repeated positions. The flag would need to be turned on (or the target square populated) only if there is a legal e.p. capture on the immediately following move.
And, don’t forget, a candidate e.p. capture is not legal if, for example, the capturing pawn is pinned vertically to its own king.
With en passant, unlike castling, there is no distinction between being “temporarily prevented” vs “permanently prevented” from playing the candidate move. If a candidate e.p. is prevented, it’s prevented, period, for the rest of the game.
For this reason, when tracking repeated positions, with en passant there is absolutely no excuse for considering a “practical” rather than a “purist” definition to determine the eventual legality of a candidate move. The situation is different with castling, where the “practical” definition is used because of the potential near-impossibility of determining the answer with the “purist” definition.
I don’t see why. If the e.p. field is populated, the last move was the pawn in front of that square.
That position, as described by that FEN, can never be repeated, and can never have occurred before. So any position with the e.p. field populated need not ever be considered for repetition purposes.
That’s what I at first thought, too. But then I realized I was overlooking something, and so is narceleb.
For most purposes, it doesn’t matter one whit whether the FEN target square is populated after every two-square pawn move, or only after those where a legal e.p. now becomes possible. The usual purpose of the various FEN flags is simply to describe the position completely, and a spurious e.p. flag is harmless.
It is another matter entirely when the FEN is used to determine whether a position has occurred previously.
Consider the opening position:
After 1.e4, the FEN populates the e.p. field (e3) because there has just been a two-square pawn move. Seems harmless, for the time being.
Now suppose the game continues 1…Nf6 2.Nc3 Ng8 3.Nb1. We arrive at the same position, but this time the e.p. field is not populated because the last move was not a two-square pawn move.
Now the game continues further, with 3…Nc6 4.Qe2 Nb8 5.Qd1. Black now claims a draw by triple occurrence. The engine looks at the FENs after white’s 1st, 3rd, and 5th moves and denies the claim, because the first FEN differs from the other two.
That’s wrong. To work properly for repetition claims, the FEN e.p. field must be re-defined so that is populated only when a legal e.p. is available.
Q.E.D.
EDIT: I posted this before I realized that jwiewel had just posted similarly.
By that do you mean “that flag is turned on only when there is a pawn in a position to capture the one that just moved two spaces”?
Or do you mean “that flag is turned on whether or not there is a pawn in a position to capture the one that just moved two spaces”?
If the former, then Good. But in your earlier dialog with jwiewel (reproduced at the end of this post) you seemed to be saying just the opposite.
If the latter, then Bad. That won’t give you correct repetition data.
But even in the former (Good) case, having a pawn “in a position to capture” is not good enough. It must also be “able to legally capture” the one that just moved two spaces.
If, for example, the potential capture is with a pawn that is pinned on the file to the player’s king, then the pawn would seem to be “in a position” to capture even though it is not “able to legally” capture:
The implementation of the above would seem to depend on what version of FEN the engine is using.
If the engine uses the software supplied by Ms. Dorothy Fen, CEO of the FEN company, you will likely see incorrect results in determining whether there has been a 3-fold occurrence. Instead, you should probably use the version supplied by the SMY company, which sets the e.p. flag only if a legal e.p. is actually available. SMY differs from FEN in this way only.
(No phone calls, please. I’m being a bit facetious here.)
Seriously, though, any engine using FEN for determining 3-fold occurrence should use the more sophisticated definition of the e.p. flag.
So when determining repetition, one must first check the first three fields: position, who has the move, and the castling rights. If those three match, then check the e.p. target square of the first position. If it is not set, you’re done. If it IS set, then you have to check whether there was a pawn in position to capture. If there is such a pawn, then it is NOT a triple (or quintuple) occurrence.
Exactly – provided, of course, that “in a position to capture” means that not only is there a pawn on one (or both) of the adjacent squares, but also that it (or one of the two) can legally capture, taking into account (for example) pins on the file(s), etc.
CONTEST !!! :: Find a position where white has just moved c2-c4, and black has pawns on both b4 and d4, yet black has no legal e.p. capture. Or prove that no such position exists.
PRIZE: :: The contest winner, in his next rated game, is allowed to move any pawn two squares forward (not just from the 2nd rank), but may be captured e.p. after doing so if the opponent has a pawn in a position to capture.
Isn’t that like REALLY easy? You probably took longer to write your “contest” than it would have taken to solve it.
I was wondering whether anyone knew officially whether e.p. is treated like castling in regards to triple occurrence. Focusing on “right to castle” vs “ability to castle” was probably a decision driven more by consistency with how computers handle positions than with the actual concept behind the rule. (The pieces are in the same positions and all moves are the same, so it really IS the same position for all practical purposes).
Why would that be relevant? If the positions of the pieces are the same, which we determine before looking at the e.p. target square, how could a pin exist when the e.p. capture is possible, but not when the e.p. capture is not possible?
(Rule 14.C mentions the RIGHT to castle or capture en passant, not the ability.)
When you enter your next tournament, please be sure to advise the TD, and your first-round opponent, of your additional rights in moving pawns two squares. You probably should show them this thread so that they can confirm.
I hope I don’t get paired against you in the first round of your next tournament.
That’s the whole point. The pin would exist in both positions. Yet, if the e.p. target square is set in the first case but not in the second, the software would erroneously report that the two positions are different, even though they are the same (e.p. is illegal in both).
And please note that the e.p. target square would, indeed, be set in the first case, but not the second, if the overly simplistic definition of “target square” is used – the one that populates the target square based only on the two-square pawn move.
All this talk about the “right” vs the “ability” to castle, or to capture e.p., simply muddies the waters. Which is the “right” and which is the “ability”? The discussion should be based on straight facts, as above, rather than on any new phrases somebody (or some USCF rule) has introduced to “clarify” (or confuse) the situation.
Not exactly. The actual wording is “if the possible moves of all the pieces are the same, including the right to castle”. That wording suggests that it’s the “possible moves” that is the criteria and they are listing the right to castle as something that might cause you to conclude that the possible moves aren’t the same. If a player can’t legally castle, one can argue that the possible moves are indeed the same and the condition is met. The majority aren’t taking that position, but a clearer rule could remove all ambiguity. I am one who believes “possible moves” means just that.
I agree. Regarding castling, FIDE has apparently made its choice based on practical considerations, and has explicitly stated this choice in its rules.
Regarding e.p., it appears that FIDE has not yet weighed in, and perhaps does not need to do so.
For both castling and e.p., a “purist” definition and a “practical” definition are possible:
Castling, “purist” definition: Castling is permanently impossible in case there does not exist a sequence of legal moves (even with both players cooperating) leading to the player castling legally.
Castling, “practical” definition: Castling is permanently impossible in case either the king or the rook has moved previously.
En passant, “purist” definition: En passant on a square is impossible for the time being in case a legal en passant capture does not presently exist on that square.
En passant, “practical” definition: En passant on a square is impossible for the time being in case the last move was not a legal two-square pawn move over that square.
Of course, the purist in all of us prefers the purist definition(s), but in the case of castling being impossible, there are serious practical issues.
Upthread, I posted the following position, white to move –
–- where, even though neither white’s king nor white’s rook has ever moved, there is no way white will ever be able to castle, even with full cooperation between the two players. (Everybody, please stare at the position for few moments to convince yourselves of this.)
A position like this could well come up in an actual tournament game, and I could easily see the arbiters (in their haste or otherwise, and even at the master level) overlooking some nuances and coming up with the wrong ruling – if they are being guided by the purist definition.
I’m sure Sam Loyd could have concocted a puzzle position requiring a dozen moves of complex retrograde analysis to establish whether or not the purist definition is satisfied.
And I could also see even the most sophisticated chess engine not being able to discover the correct answer, because the engine would have to generate a huge number of possible game continuations, not just for 1 move, but for the rest of the game.
FIDE doesn’t want its rules to mask hidden complications, so in the case of castling being impossible it has chosen the practical definition over the purist definition. FIDE has done the same with its dead-position rule – K+N vs K+N is not an automatic draw because help-mates are possible.
Now, since FIDE has chosen the practical over the purist in the case of castling being impossible, should it do the same for en passant being impossible, just to be consistent?
Those who decry consistency often quote the phrase “Consistency is the hobgoblin of small minds.” I hate it when people say that, because they invariably omit the first word of the phrase: “A foolish consistency is the hobgoblin of small minds.”
Now, would consistency in this case be foolish? In my opinion,YES.
In the case of en passant, the problems with the purist definition are minor, and can be overcome just by being careful about the way the e.p. flag (or target square) is defined.
Also, the situation with e.p. is easier than with castling for an additional reason: with e.p., there is no distinction between a move being “permanently impossible” vs “temporarily impossible”. Those ugly ambiguous phrases involving “right” vs “ability” never enter the picture.
So, my FIDE colleagues, when deciding whether a repetition of position has occurred, I implore you to keep on explicitly supporting, in your written rules, the practical definition for castling being possible, while holding on to the purist definition for en passant being possible. “Consistency” is not necessary. To keep the purist version of en passant, all that’s needed is to remain silent. Then it will be the default.
Further, in the case of en passant being impossible, I would prefer to rename the “practical definition” as the “simplistic definition” or even the “incorrect definition”. “Practical” is just too generous.
By contrast, with castling being impossible, “practical” is fine.
Geez, I hope I get some +1 markings on this one. But I do realize it’s a long post.
I am laughing hysterically at the idea of arguing for “practical” vs “purist” while using the diagrammed position as an illustration. There is zero practical chance of said position appearing three times in a game.
My point was, there may be positions where it is difficult, perhaps near-impossible, to determine whether future castling is possible. And, who knows, some of these positions may even be plausible, repetition-wise.
I’m sorry if I’m not exactly Sam Loyd. Perhaps somebody out there could come up with a better example.
You abbreviated that quote. It says, “including the right to castle or to capture a pawn en passant.”
So, the question is, HOW WOULD YOU RULE?
Given a draw claim in which the first position occurred right after c2-c4, with a Black pawn on d4 which is absolutely pinned (say White Queen on d1 and Black King on d8 and nothing else in between), do you uphold the draw claim?
The options are:
YES, one does not have the RIGHT to make a move that puts his King in check, so the possible moves are the same at all three points in the game.
NO, although illegal, Black had the RIGHT to play …dxc3. (This has the advantage that play continues.)
Upon consideration, I would uphold the draw claim.
White Kh2, Qh3, Bd2, Pg2, Pf3, Ph5. Black Kg5, Qd7, Bc6, Pf4, Ph6.
1… QxQ+, 2. KxQ Bd7+, 3. g4(*) Bc6, 4. Kg2 Bd7, 5. Kh3() Bc6, 6. Kg2 Bd7, 7. Kh3()
White’s 4th and 6th moves were the only way to avoid dropping the f-pawn. White’s 5th and 7th moves were the only way to avoid Black trading a bishop for White’s 3 pawns (initially overlooked by white). White’s original plan was to play Kg2 and win either the f or h pawns by playing Bb4 and threatening d6 (winning f4), e7 (mate) and f8 (winning h6) and he missed that it was still a viable threat on move 4.
