Dead reckoning puzzles

The comments relate to this diagram:

Whether dead reckoning applies or not, retro-analysis reveals that white may have had, or had, castling right in the proof games for the diagram. By the castling convention in the Codex white therefore has permission to castle and 0-0-0 checkmates in 1 move.

This is not actually a dead-reckoning problem because the castling convention washes away the distinction between castling right and castling permission. It would have been a dead-reckoning problem if the castling convention woud have read "castling is not allowed unless it is provable in all proof games" - like the e.p. convention. The diagram actually proves castling right when the dead reckoning convention is in force.

What happened? Jumped from July 2015 to November 2019. Four and a half years for . . .

To show that I am not here to debunk every DR puzzle in the universe, here is a problem based on the same idea as the one in my last post but with a meaningful DR application.

Answer: Yes! White can take e.p. and checkmate on the next move. Why? Black only has 3 last moves (a) Ph6xg5, (b) Kg7xh7 and (c) Pg7-g5. However "(a)" and "(b)" are impossible as the positions are already dead due to inevitable draw. What remains is "(c)" permitting the e.p. solution given.

Please verify that the solution is in line with the last paragraph of my previous comment where I indicated that you can do with e.p. what you cannot do with castling because the two conventions differ.

Note1: I recall JoachimJo posted a pawn ending with similar logic 2 years ago but it appears that chess.com has removed many posts in long threads and I cannot locate it any more.

Note2: Remember that by today's conventions, this would still be a silly problem if I had not added the retro problem type to the stipulation. See my forelast post!

Since 2001, I have argued that castling & e.p. conventions do not engage until all retro/forward reasoning using the Laws including A1.3 is otherwise exhausted. These conventions only grant "permission": what we are talking about here is legality, by establishing facts about the history of the game which are sufficient to determine whether these conditional moves are or are not viable. Almost all DP problems involving the castling & e.p. conventions function in this way. It is certainly possible to combine the DP rule with the conventions in compositions, but these would not imply (in my reading) the DP rule having any visibility of the action of the conventions. I know that Arisktotle has his heart set on some grand unification of all of chess problems: retro, fairy and otherwise, but the basic idea that conventions are secondary to the Laws remains in my interpretation, which I believe is self-consistent

@anselan: You inverted the point I made. The 2 last diagrams are actually analytically similar to "the castling DP" problem you messaged to me some weeks ago. There is no issue here of "DP depending on e.p./castling" but the opposite just as in your composition "DP proving castling/e.p. right". The point I made was that from a solving perspective anyone not knowing or not thinking of the DP rule would solve the castling problem anyway because castling is permitted. That is different for the e.p. version which someone not knowing DP would find unsolvable. There is nothing here which we disagree about I think.

Much more relevant of course are my notes on the DP diagram in #7 which perfectly illustrates the issues I have with the "meta-choice" concepts - in this case between retro and endgame.

There remains a little disagreement. Earlier in this thread you wrote: “This is not actually a dead-reckoning problem because the castling convention washes away the distinction between castling right and castling permission”

Well it is a DP problem, because we only engage the conventions when we have need of them. Someone unaware of (or defiantly opposed to!) the DP rule could reach the same answer for different reasons, but that is a separate issue. These conventions are secondary to the rules.

I like the idea that a chess problem relates to a game like "a faithful film of a book". There should be an absolute minimum of adaptation from the rules. In this spirit, let's deconstruct the castling convention. (You made the same point in a different way.)

"Castling is permitted unless it can be proved that it is not permissible". What does this mean? It's conflating a convention with half a rule. There are three cases:
(1) Castling provably permissible => *can* castle (by rule & convention)
(2) Castling provably impermissible => *cannot* castle (by rule & I think also convention, although usage of "unless" to mean "unlesss" isn't universal. However this question of interpretation turns out not to matter here.)
(3) Castling cannot be proved => *can* castle (by convention only)
Key argument: there is no point reiterating a rule in a convention - all we need is (3):
"If it cannot be proved whether the relevant castling right remains, castling is permitted."

And when properly written in this way, the castling convention therefore interacts harmoniously with the DP rule. A very similar issue occurs with the e.p. convention, and the fix would be: "If it cannot be proved whether the last move was the pawn double step, en passant is not permitted."

There is no point losing perfectly good Design Space just because of current sloppy writing of a convention, when the objective is clear. In anticipation of these conventions being rewritten 15 years from now maybe, I will assume that the convention states only what it needs to.

The one in #23 is fun. I would put wSe8 on d7. The analogy to castling is *failing* to e.p.

Yep, I know we have this "litlle" disagreement. My comment approached the subject as a solver assignment not as a purely theoretical analysis. Had you been in the discussion at that point, I would have written my comment differently. I always prefer to see the item under discussion have an observable effect on the outcome of the assignment - but I agree there is an extra double negation in the equation where a specialist might go wrong!

I don't think Se8 could be on d7 in #23 because blacks last move might have been ...Kg7xh7. Then, it was only an example problem because I couldn't find JoachimJo's brilliant endgame version. Did you ever see it?

If Black just played Kg7xh7 it would have been forced, as both f8 & f6 are protected by wS. JoachimJo rings a bell - think it's https://pdb.dieschwalbe.de/P1348737

You're right it's very nice. Good not to have wK beside bP on 5th rank

I agree good for "item under discussion to have an observable effect". However, there's too much great design space in pure DP to avoid it just because the castling/e.p. conventions have been badly drafted. Also the logical thinking of the different cases listed above is a good exercise to put people through. I rewrote the e.p. page in Retro Corner many years ago, and some people were surprised that what they had thought were examples of e.p. convention were just simple cases of e.p. rule playing out

Yep again! I am losing my touch. That would actually make it sort of a nice problem!

That's probably the same as I have been thinking for a long time. There is no point in stating that an e.p. move can be executed when legal in all proof games, because as a meta-convention you can always execute anything that is legal in all proof games. It should have been inverted to identify the situations where you are not permitted to execute an e.p. move!

I think each convention should only say what you should do in case of unprovability. The same should apply to: whose move, 3Rep history, 3Rep claim, 50 move history, 50 move claim

The whole subject of retro-logics only consists of those cases with the connotation that unprovability does not include disprovability! 

I must go to bed - work day tmr. I changed the text you just quoted to add an "only". The Sd7 suffers from 1.Sf8# so you haven't lost your touch but Be7 looks ok. Really want to find the best forward play with this kind of arrangement

Thanks! And I only quick composed #23 hoping to replace it with JoachimJo's endgame once I found it. I'll check out your link and will spend a few more minutes on meditating #23.

Sleep well!

Edit: Followed your suggestion and edited post #23. Sort of decent problem with your modification. Also like the 5 x 4 artistic rectangle in the position.

Edit: By the way, can't find Joachim's diagram with your link. I get 31 diagrams but there is no pawn ending amongst them. Also, don't know JoachimJo's real name.

Have fixed the link for Joachim Iglesias' study.

I like the new problem - I think the stipulation should be just #2.

wBd4 only contributes to the forward play but in fact it is contributing passively to the retro too in that it doesn't let the looming pat to be broken (except by e.p.).

Do you honestly still think that the problem in #20 is unsound, when the only issue is sloppy writing in the castling convention by two guys who were not native speakers of English? I would appreciate if you can tone down / reverse the judgment of #20 & #23.  And I won't escalate to bold if you don't

(By the way I think "unless" in the Codex is always intended to mean "unlesss")

Info for the problems posted earlier in this thread:
Andrew Buchanan 1 Retros mailing list 24/01/2001 Whose move?
Andrew Buchanan R0093 StrateGems 18 04-06/2002 WTM. Last move?
Ronald Turnbull V StrateGems 10/2001 WTM. Last two single moves?

So all these are retros simply because of the stipulations. Hence Arisktotle's wise Disclaimer ("obviously I have assumed the puzzle was posted as originally published") applies. I would appreciate if he can edit post #20 accordingly.

Re his Note 2 in "#23", there is the broader question of whether a purely forward stipulation could allow the problem to be classed as retro. However because our minds think alike (mon semblable, mon frere) Arisktotle is also alert to the "circular argument" risk - and indeed used the same term.

The term retro is deliberately left undefined in the codex today, so I feel I am ok to use sloppy meta-reasoning to say: if the problem on inspection has retrograde analysis content, then it's a retro. This is easy enough to see in practice. If the problem has no retrograde analysis content, then the stipulation should say: "Dead Position rule applies". This indicates that DP is not a fairy condition. (In PDB, I have marked these with the non-pejorative keyword "golden age" to indicate they were impacted by rule or convention change. In many cases, those problems were fixed by adding retro content, which substantially enriched them, so the convention change was positive.)

For me, the acid test as to whether a problem is retro is: "Suppose the initial position in the game was the problem diagram. Is the solution still the same?" What exceptions are there, where this test does not work as we would like?

How can a solver recognize the *context*? This is a different question: I don't think a solver would be concerned by the "circular argument", but doesn't want an unfair trick, and nor do I. In almost all cases, however, I think the context is there: whether it's a DP article, or keyword in database entry, or reprint in forum like this. In any case, the problems are simple enough that hours of solving time would not be fruitlessly invested. In the long run, I hope the notoriety of this idea evolves to a simple awareness that this is an item in the composer's toolbox, without having to resort to clumsy additions to the stipulation.

Ke8x/-d8 is the try, for reasons given by Arisktotle