In the thread Draw by position occurring 3 times, there seems to be general consensus (maybe not quite unanimous) that two apparently identical positions are not considered the same if the right to castle has been lost between the two occurrences.
The above remains true even in the case where castling is temporarily impossible in both positions, but permanently impossible only in the second position.
For example:
If the game now continues 5.Rh2 Be7 6.Rh1 Bf8, the resulting position is not considered the same as the diagrammed position because, in the second position, white’s (eventual) right to castle has been lost.
But, even if we (mostly) agree on the above, there can still be sticky-wicket situations. See my next post.
Here are two possible definitions of “castling permanently illegal”. For simplicity let’s consider only white kingside castling:
A. There does not exist a sequence of legal moves leading to white castling kingside.
B. Either the Ke1 and/or the Rh1 has moved previously.
Thought questions:
Are the above two definitions equivalent?
If not, does either definition imply the other?
Which definition should be used in the USCF and FIDE rules governing triple occurrence?
Which definition do chess engines (those that make or allow draw claims) use?
Which definition would you use if you were writing a chess engine?
Which definition should a chess engine use?
In the following position (white to move, neither the Ke1 nor the Rh1 have ever moved), is white kingside castling temporarily illegal, or permanently illegal?
Examine the position carefully before responding. All entries welcome!
Edited to correct definition B. I had it backwards.
That second one may not be the best example. Unless I’m missing the obvious, there is no way for there to be triple occurrence of position until after the point becomes moot. After b6, Kxb6, white is forced to abandon castling altogether.
Regardless I understand what you’re getting at. I’m probably too new at this to have a proper opinion, but my suggestion is only when the “right to castle” has been permanently lost rather than temporarily lost should this change the position. The “right to castle” is only lost when the king or both rooks have moved. This is a change in the fundamental state of the game, while everything else is just logical and temporary.
Isn’t this how FEN notation defines it? My guess (and I don’t know for sure either way) is that engines would treat it the same way.
I’d say A is the purist definition, B is the practical definition. It would seem next to impossible for a chess engine to use definition A to determine the validity of a draw claim – until the day arrives when computers have completely solved chess.
Definition A is a chess lawyer’s dream, but in my opinion it is simply wrong. In my view, until the king or rook has moved, castling is not permanently illegal.
However, I question it is possible to construct a position where the nit-picky subtlety of A could come into play but a three-fold repetition based on that position is possible. Can you construct such a position?
No, because the conditions are reversed (i.e., A tries to define when castling is permanently illegal, while B tries to define when it is not permanently illegal). But if B is false, this would imply that A is true. But if A is true, this does not necessarily imply that B must be false.
The FIDE rules clearly indicate that B is correct. And I believe that the USCF rules imply the same interpretation, though I agree that the wording does not make this clear. Adopting A, instead, would make application of the rule less straightforward, since it would require that the TD evaluate the board position and determine all possible lines of play. And I see nothing that would be gained by changing the rules to use A rather than B.
In the above position white’s castling is only temporarily illegal. It both players lose track of whose move it is and black plays Rd8 then white can legally castle.
If you can pick nits then so can I.
Is there really any difference between “having to give up the right to castle because only the King and Rook have legal moves” (the case you’re bringing up) and “having to give up the right to castle because you’re in check and only a King move is legal”? Because the FIDE rules rather explicitly and the USCF rules (not quite so) explicitly say that in the latter situation, the position is different than in will be if the same position is repeated a few moves later.
Suppose, in the above position (with neither Ke1 nor Rh1 having moved previously), the following moves ensue:
1.Rh2 Rf7 2.Rh1 Rf8 3.Rh3 Rf6 4.Rh1
– and now black writes 4…Rf8 and claims a draw by triple occurrence.
The correct ruling would now depend on the definition of “castling permanently impossible”. By definition A, there has been a triple occurrence. By definition B, there has not.
So the FIDE and USCF rules are using definition B, then.
My view is that definition A is more satisfactory when it comes to making a ruling, but that definition B may be the only way a chess engine could implement the rule. As I said before, it’s the purist definition vs the practical definition.
I disagree that A is even a good purist definition. Being in a 60 mph car five feet from a brick wall does not mean that I have already crashed. There is a difference between “you have lost the right to castle” and “it is certain that you will lose the right to castle”.
It would be easy, through appropriate triangulation by one of the rooks and/or one of the kings, to create a position where white claims.
I don’t understand that analogy at all.
This would not be the only rule where a future possibility or impossibility determines the outcome. For example, look at the FIDE dead-position rule, or its USCF equivalent, 14D. That rule declares a game drawn if there is no sequence of legal moves leading to either player checkmating the other.
Nice problem! White’s ability to castle is permanently illegal. White can move either the king or the rook but not castle. If white does so castling becomes permanently illegal. White can move the pawn with check. After this check black can capture creating the position where castling will become permanently illegal or move away and the blockaded pawn fixes thing for white. Rule B is the best way to program and explain the rule because it helps to partly clarify Rule A. I had to edit my answer when I thought a little about computers.
The point, of course, is when has repetition been achieved? After 3 moves or after 4, since castling is still logically (if not practically) available? I think I’d have to say after 4. Fortunately this situation is extremely rare. I don’t relish having to explain this one.