Unbelievable to me

I've just edited the diagram in post #86 reflect the changes discussed in posts #88-94. It's probably high time I moved on to something from 2015, so here goes nothing:

...But first, a word of "caution" () - the presentation of the composition that follows is not at all conventional. When I came up with this 2 years ago, I considered its presentation to be more or less optimal, but knowledge of standards and conventions has a funny way of altering one's perception of how "good" or "bad" certain stipulations really are. 

However, devising a more acceptable objective for this problem would still require me to wait even longer before posting it (or choose another puzzle from the 2015 archive, but I would definitely prefer not to do that!), so I'm "laying it out there" as is, and reserve the right to make changes later (especially if the criticism is even more intense than expected ).

Anyway... there are two parts, plus an extra condition which was intended to link the positions in a nontrivial way (to prevent them from becoming two separate puzzles with coincidentally similar appearances, as in my problem from 2011). Here is the original version, copied word for word:

Add a unit (edited for clarity) to one of the diagrams below, and then identify the promotion(s) that occurred prior to each position shown.

Without having made any attempt to solve it, my first impression is that this is a contrived stipulation indeed. You are not shocked because, with your born-again perception, you did expect this assessment. At the same time, from the many discussions, you will have concluded that there are no absolute rules about stipulations and you are ultimately free to choose them and create your own following! And of course, there is still the rethink.

I do however have two questions on clarity. Is the adding of the unit supposed to be unique on a unique square? The answer would be clear if adding a piece was the total stipulation but as a stepping stone to the promotion question, it is kind of vague. Also, "add piece" could/could-not be a pawn here? The questions may prove to be irrelevant but they arose before attacking the problem.

#86 looks great now! I hope you checked it thoroughly since I triggered more on the promising idea than on its complete validation.

This is one of those problem scenarios (pun very much intended ) where I still love the content, but wish I'd been more informed when translating it into an actual composition. I fear the "rethink" may demand changes of a magnitude I'm not really comfortable with, but perhaps your input (or anselan's) will help prevent the dreaded "complete overhaul" from becoming necessary.

Well (as you obviously know, being a professional composer and all ), from a composer's perspective, I would certainly like to think the full solution is unique... including the placement of the extra piece! From a solver's perspective, uniqueness was "supposed to be" much less of a concern, since that sort of knowledge tends to aid a great deal in solving.

I've always strongly disliked the distinction that is often made between "pieces" and "pawns", and should have realized it would lead to ambiguity here. As you can probably gather from the previous sentence, my preferred definition of "piece" is the one which refers to any of the 6 standard piece types in the game of chess (as in "7-piece endgame tablebases" or "each side begins with 16 pieces").

Thanks! 

At this point, I'm more convinced than ever that it takes at least 2 or 3 people to check anything thoroughly... especially when one of them is the original composer. However, just like I did with my 2011 puzzle, I did search around for additional ways to cook the problem before accepting the fact that I would need to rely partly on others' expertise () in order to cover absolutely everything.

One of the disadvantages of textfull stipulations is that language vagueness plays a bigger role. Sometimes "piece" does indeed mean "not a pawn". I personally prefer the descriptions "unit" and "man" as in "7-men endgames". Don't think that the chess world is unified on this though.

Just checked brobeckers pdf which is all about adding material to get legal positions. He uses "add unit" to cover this action.

I will test #86 once more!

Just FYI: when I use the app on my iPhone, I can't read any caption stuck on a diagram.

This identification task I think is implied for any retro problem or retro-forward problem. The knowledgeable solver generally will be looking for any well-defined sequence of moves back from the diagram. If you want something more definite, I suggest "Last moves?"

I always enter everything stipulatory as "event" information and ignore captions. Never had any complaints about that.

@#102: It's a good piece of structural content, requiring thorough investigation of the available retraction strategies! My opinion is that having a pair of diagrams is hard to justify, but the stipulation is reducible to one diagram without losing significant content.

The following analysis doesn't prove that "+wPd3" is the solution but it shows that it is the only addition that separates diagram 1 from diagram 2 and therefore must be the solution if the problem is correct.

Comparing the possible unwinds delivers the conclusion that both diagrams (after adding a unit) are retro-analytically the same in all aspects except the possible bishop promotion on a1 in diagram 1. Diagram 2 has no plusses and will always lose to diagram one in the unwind except when there are too many promotion choices in diagram 1. But there never are since opening up the cluster always releases several different units to unpromote on a1/a8 which can't be the solution. Which leaves the unpromotion of the bishop on a1 as the only way to separate diagram 1 from diagram 2. Thus: +wPd3.

One can argue that the preceding analysis is precisely what makes the twin presentation attractive but I doubt that is the case since almost the same bits of analysis must be made if there is only a diagram 1.

---------------------------------

Both the twin and the single diagram setting have the disadvantage that there is no temptation to unpromote on a8 because there are no white pieces available to unpromote. If the solver fails to see the release option with the h5-h4 retraction, he will still find the correct solution because it's all that is left. I would therefore advocate replacing (Bd6) for instance by (Rd6). That would demand the solver to see the h5-h4 retraction work, since only that cooks the rooks unpromotion! And makes sure that he sees the h5-h4 retraction fail in the solution line with +wPd3.

Consider it fixed.

I'd very much appreciate that. 

And btw, anyone else following the thread is more than welcome to take a look at it, too - the more, the merrier!

It's good you mentioned this, as otherwise I'd never have known (since I don't have the app). At the moment, I'm too lazy to go back and edit past diagrams (maybe when I have a little more time), but from now on I'll make an effort to avoid using diagram captions. 

Totally lost me. How is "Last moves?" any different from what I used, apart from being shorter and a bit more open-ended? Also, why exactly would an answer such as "The last move was gxf5+" be wrong, with your suggested stipulation in place of the current one?

Rats - I really screwed this one up, big time. The worst part is that the cook was incredibly trivial to repair, if only I'd seen it. 

I'm editing both diagrams as we speak. The quality of the resulting problem takes a minor hit, but (mercifully) most of the content remains intact... I think. 

The "cook" was a 50% cook (only when adding a bP, not with a wP) and I thought you intended it. The solution was still correct.

In essence, the comment I made on the separation of dia1 and dia2 remains the same. Even more so now that some retractions options disappeared.

Even if I'd specified the exact square where the extra unit must be placed (already unacceptable, imo), the diagram without the extra unit would've had 4 solutions. That rips the entire problem to shreds, in my book.

Sadly, those retraction options were (seemingly) more destructive than you realized, even though you were the first to notice them. Their disappearance reduces the depth of content quite significantly, but that's a price I'm willing to pay in exchange for soundness. Every component of the solution is now unique, and after solving the new version, you'll see why it can't easily be reduced to a single-diagram problem.

... And I just realized it's still cooked. Yeah - time to scrap this one forever. I'll post something better (also from 2015) tomorrow.

Really sorry about this; it was a truly gigantic waste of time for everyone involved.

It's OK. I just saw the extra pawn can be on f7 besides the d-file which may be repairable. Are there other spots? I am not convinced it is irrepairably cooked as yet. Even though I can't know what you originally had in mind which apparently never worked.

Note: I only ever tried to define the promotions in the diagram where the unit was added, not in the other one! Misread the stip apparently. However, if the 'other' diagram is a lost cause, it is a good excuse to concentrate on getting the one with the extra unit right.

The intended solution was to add bPf7 to the 2nd diagram, so that a1=B occurred in the 1st diagram and a8=Q occurred in the 2nd. I thought removing the h5-h4 retraction option was enough to save it, but I was wrong - the promotion on a8 could have been to anything, not just a queen. So the problem can only be salvaged by discarding the 2nd diagram altogether, but of course the 1st diagram alone is not much of a puzzle... especially once the original h5-h4 cook is dealt with. Your suggested fix (adding wPd3 and replacing wBd6 with wRd6) would be a clear improvement over what I did (moving bLSB to h5), but still gives away an awful lot. And a new stip such as "Add a White unit, then identify the promotion that occurred" would merely be grasping at straws.

Yeah, I just guessed that adding Pf7 was supposedly the solution to the 2nd one and saw the consequences. I'll continue to ponder on this for a while and then (probably) give up.

Btw, I didn't want to suggest adding wPd3 in the remaining diagram but still ask to add a unit anywhere for an unambiguous promotion. wPd3 is then the only option even with (Rd6) on the board - I hope. Still, I understand that giving up the 2nd solution is a serious blow to the design.

I suppose there are a number of ways to repair the problem but it depends on what you want to keep. Since I am not that excited about the twin diagrams, I made one with a twin solution:

I do believe more can be done with these ideas but this is about as much time I want to spend on it.

^ Well done!! The caged rook on e5 is very much in the spirit of what was intended (after seeing how the original rook cage could be busted, I seriously doubted whether any version of it could be made to work), and meanwhile, you managed to incorporate a brand-new idea involving the "free" bishops!

I'm sure anselan will want to debate whether you are now the sole author or a joint author, but regardless of how that question is answered, I think the above problem is very, very nice. 

Thx, Cobra! I tried to stay close to your patterns and solutions in order to have you look on it kindly as your own biological offspring though raised in a different family.

I have my reasons for not being completely satisfied with this version. I actually tried to create an illegal diagram position where the pawn on d3 would provide a critical tempo in liaison with the unpromotion on a1 plus retraction to a4. I Failed, though I am not sure it is impossible. Edit: achieved it but at a price; if you want it I'll post the diagram.

We composers hate free units . We believe all units must earn the right to appear in a diagram on the squares they do!

Missed this. I'd actually like to see how you managed it, as it takes a lot of tempi for Black to make all the necessary retractions. Avoiding the use of a free White unit seems impossible, but maybe that's only because I haven't thought much about it.