A.Buchanan: The point I think is that the en passant is a tempo move. Otherwise wBf2 serves no purpose. How about sBh5 instead of sT? (2021-10-19)
Henrik Juel: C+ Popeye 4.61 (2021-10-19, hidden) more ...
comment

paul: See P0003212 as version. (2011-08-06)
Anton Baumann: C+?? Die Aufgabe ist eindeutig inkorrekt! (2021-01-27)
James Malcom: Fixed. (2021-01-27)
comment

Henrik Juel: The missing black man is dark-squared, so last move was not fxe4. In the solution Td6 should read Txd6. There seems to be a 'variation': 0... dxe3ep 1.0-0-0 Le4 2.Txd6 Sg4 3.Tf6#, so maybe the stipulation should have 1.1;2.1;1.1 added. (2004-09-23)
A.Buchanan: Well spotted Henrik. Your kindness is legendary, but in the harsh world of helpmates, this counts as a cooked I believe. In fact, it's the 2.Lf5/3.Sf3 version which is the cook. Yours is the actual solution! This is because 3.Sf3 would check if White hadn't castled, and we want the castling to be because of the AP condition only, not for 2 reasons, which one might term a "logical dual". The problem can be trivially fixed by sticking sSe5 on f6. I have no idea whether it's a typo or a cook: WinChloe has the same diagram, and the same confusion between cook & intended solution. The fact that AP does not appear in the stipulation might be a factor. WinChloe as far as I can see has no concept of AP as a keyword, and just picks it up based on a substring of the stipulation. In other compositions, T.Petrovic wasn't shy of flashing the AP acronym in the stipulation, but as usual the detective work peters out in the absence of primary records. (2020-12-08)
A.Buchanan: But maybe I'm being hasty. Perhaps the intention *is* to have two variations? Can anyone get at relevant issues of Problem, please? Or provide me with an online link to the relevant image files, and I'll check. What do you think Henrik? (2020-12-08, hidden) more ...
comment

paul: White captures was axb and g2xh3, so the retro move c3xb4 is not possible. If b3-b4, the retro check is not justified. So last move was b2-b4 (preceded by Rc3-a3). (2011-08-06)
A.Buchanan: There's a lot to unpack here. Firstly, the intended solution has a definite transposition dual in the order of Black's 2nd & 3rd moves. Second, the pawn capture count is wrong. What happened to the original wBh, given sBh2 never left the h-file? Thirdly, and not a showstopper, White's only solution involves castling: so that's a logical dual. (2020-12-08)
A.Buchanan: There's a lot to unpack here. Firstly, the intended solution has a transposition dual in the order of Black's 2nd & 3rd moves. Second, the pawn capture count is out. What happened to wBh, given sBh2 never left the h-file? Thirdly, White's only solution involves castling: it is not required to retro-justify the ep which was certainly legal. So that's a logical dual. (2020-12-08, hidden)
A.Buchanan: Ok looking at this again. First the keyword was wrong - this isn’t an AP problem at all. Second wBh can have captured bBg. But the transposition dual is a showstopper. We can shift bK to g1 as wBh can’t retract without blocking wLf1 exit. But that does nothing - we really want bKh1 already, but then wKf1-e1 was possible as last move. Surely this is fixable as AP, but that wasn’t the author’s intention (2021-05-01, hidden) more ...
comment

Henrik Juel: Without sPf7 the set play would be cooked: 1.gxh8D dxc3 2.Dg7#. (2004-09-13, hidden)
A.Buchanan: Yes +sBf7 corrects an earlier cook in the set play (2021-10-21, hidden)
A.Buchanan: See P0003286. What was published? (2021-10-22, hidden) more ...
comment

Henrik Juel: 0... fxe3ep 1.d8S Kxd5 2.0-0-0#. Not -1.f2? and Lg1 is caught. (2004-09-16)
Vaclav Kotesovec: Similar problems should not be labeled as "C+". Such a designation is only acceptable if the entire analysis was performed by a computer program. (2023-08-03)
Henrik Juel: In principle I agree, Vaclav
But PDB does not (yet) allow HC+, so I find it acceptable to use the C+ label, when you also tell the whole story after 'Computer test:' below (2023-08-03)
A.Buchanan: Hopefully Gerd will have more time at some point, and can expand the functionality in this and other areas. In the meantime, engine solving of conventional retros including AP, is in its infancy. Retractor 2 has some effectiveness, but is still basic. And there is nothing that yet grasps the intricacies of castling/ep etc. However AP problems do often contain considerable forward chess, and the C+ tag is very useful to filter out those that have already been solved forwardly, without pretending that these are in any sense fully solved (2023-08-04)
Ladislav Packa: I know the definition of AP, but I don't understand the logic behind it. The move 1.fxe3 e.p. proves that White CAN castling. But the solution (2. ...Rd1#?) claims that castling is MANDATORY. From my point of view, AP is correct when only castling is necessary for the solution and the Rook move would be a dual. (2023-08-04)
A.Buchanan: @Ladislav: I am not sure how to help you. Maybe you can read this page from Retro Corner: https://www.janko.at/Retros/Glossary/APosteriori.htm (2023-08-05)
Ladislav Packa: Andrew, what should the article help me with? I quote the final sentence:
Some people still oppose this rule and argue that it should certainly not be the default convention. (2023-08-05)
Henrik Juel: You could view it this way, Ladislav
h#2 means that it is Black to move, so White made last move
What was last move? A little analysis shows just three possibilities: f2-f4, move by Ta1, or move by Ke1
So normally we cannot assume that last move was f2-f4
But if White can castle, then the last move was f2-f4
So if we could start with 0... 0-0-0, then 1.fxe3ep would be legitimate
AP says that you are allowed to reverse the sequence of events; first do the ep capture, then later legitimize it by castling
Was this helpful? (2023-08-05)
Ladislav Packa: Henrik, you don't have to explain that to me. I've done a few AP issues myself, like P1348357. But that doesn't mean I agree with AP's logic. I already wrote it - the term "you can castle" is applied as "you must castle". But these are only problems where, in addition to castling, the Rook move can also be used, I consider that a dual.
From that point of view, the P1000662 issue is perfectly fine for me. (2023-08-05)
Joost de Heer: AP: By castling, you prove a posteriori that the ep-capture was not just a try but the actual solution. Without castling, the ep-solution just is that: a try.
So: Try 1. fe3 ep e8=S 2. Kd5 Rd1 - but ep capture not allowed, as there is no proof that f2-f4 must've been the last move.
Solution 1. fe3 ep e8=S 2. Kd5 OOO - Now the ep capture was justified because white castled, thereby proving that the last move before the diagram position indeed was f2-f4. (2023-08-06)
Joost de Heer: See e.g. P1052919 : The try is an ep capture which is unjustified. (2023-08-06)
A.Buchanan: OK Ladislav: I think I get your point. If one solution with castling justifies the e.p., then based on that certainty, why shouldn't an alternative solution with no castling *then* be allowed as well? There are problems in which one twin shows 0-0-0 and the other shows 0-0. Each is based on the other in a similar way, so the idea of dependency is not new. Why are we not allowed to add other "parasitic" solutions as well? Why can we only have the "paying" solutions? We can't say that we are restricted to one solution: that's not the way chess problems operate! And this is just in the help world - in the adversarial world it might get even more complicated. Is this your issue, Ladislav? (2023-08-07)
Ladislav Packa: I don't want to unnecessarily prolong this discussion. However, I will add one more note: in this position, white castling is also possible without e.p. in Black's 1st move. If B1 were an indifferent move, then white can 1...0-0-0! The Codex of Chess Composition writes about it in Article 16 (1):
Castling convention. Casting is permitted unless it can be proven that it is not permissible.
In our case, 0-0-0 is possible because White's last move exists - e2-e4! It does not matter if it is this move or some a2-a3, both moves are equivalent. EP does not prove the possibility of casting, it would be legal even without it. (2023-08-07)
A.Buchanan: Ladislav was what I wrote your issue pls? Y/N :-) (2023-08-07)
Ladislav Packa: I have no problem, I'm just expressing my own opinion about the AP convention. (2023-08-07)
A.Buchanan: OK cos I think the point I raised is a real one that should be addressed by theory some day. Clearly from the nice problem that you composed Ladislav you understand the mechanics very well. From a justification perspective it's all a bit iffy, but that's why it's controversial. Under RS it's really the only way one can end up actually eping, and it's proved compositionally fertile. So that's enough to justify (2023-08-08)
Ladislav Packa: No need to apologize. This is a normal discussion with different views on the issue. Maybe it will come to some conclusion.
I just want to point out the fact that it is not the e.p. that authorizes castling, because according to the Codex castling is possible regardless of the e.p. It's the exact opposite: castling authorizes the possibility of e.p. (2023-08-08)
Joost de Heer: "I just want to point out the fact that it is not the e.p. that authorizes castling, because according to the Codex castling is possible regardless of the e.p. It's the exact opposite: castling authorizes the possibility of e.p."
You misinterpret AP. The e.p. capture does not authorize castling, castling provides a justification later on (hence the 'a posteriori') for the legality of ep.
Usually, for ep justification you need to examine all game trees that lead to the diagram, and only if all game trees end with the double-step, then ep is allowed.
With AP, you examine all the game trees including the actual play. If all those game trees have as last move before the diagram position the double step, then ep is possible.
In this case, if white doesn't castle, then there are game trees which don't have as last move the double step, and therefore AP logic dictates that the ep capture was illegal. However, all game trees which lead to the diagram and which have castling in the actual play have as last move before the diagram position the double step, hence AP dictates that the ep capture is legal. (2023-08-09)
Ladislav Packa: Joost: A simple question - is white allowed to castle after any 1st move by black (except e.p.)? (2023-08-09)
Joost de Heer: Of course he is. AP only is used to combine the ep justification with castling, not the castling right per se. (2023-08-09)
A.Buchanan: I agree with both. Also "cooked" is closer to "unsound". And we choose not to have "C+ cooked", which might legitimately mean "C-" (2023-08-04, hidden) more ...
comment

A.Buchanan: WinChloe doesn't think this problem is cooked at all. It says there are two full solutions, with "Echange de place de pièces noires entre les positions finales" i.e. Kb7 La8 vs Ka8 Lb7. WinChloe does not cover the (single) set play. Must the number of set plays normally match the number of full solutions? I will mark this problem HC+ uncooked. (2021-10-31)
comment

James Malcom: This is a most excellent problem! En passant as the key and a double pickaninny in a short s#2 is surely no easy feat, I'd say. Now, question: Why can't dxc5 be Black's last move? The illegal retraction od b6-b7 is already quite obvious. (2020-10-24)
A.Buchanan: R: 1. cxb6/dxc5? both imply 3 pawn captures by Bl (2020-10-24)
A.Buchanan: I agree this is a beautifully integrated composition (2020-10-24)
Henrik Juel: Probably the man on e8 should be a wL. 1.dxc6ep .. 2... Sc7#. Double pickanninny. (2004-09-23, hidden)
Henrik Juel: C+ Popeye 4.61 after analysis shows that last move was c7-c5 (2020-10-24, hidden) more ...
comment

A.Buchanan: Suppose White: dxexf,cxdxe,bxc=,a| Black: fxg. This means there is one capture by white unaccounted for. So can’t White have just played e.g. QxS? What am I missing? (2022-03-22)
Mario Richter: Regarding the "can’t White have just played e.g. QxS?-question", the following considerations may be helpful:
1. black Pawn f3 x Yg2 can only be retracted after wPf2 has returned home.
2. If at this moment the black K is still on g1, Y cannot be a white rook.
3. Assuming that W can still castle, wRg5 must be a promoted Piece, since wKe1+wPf2+wPg3+wPh2 form a cage from which the originalwRh1 could not escape ...

I think the same considerations can also be useful by answering the question about +sBc6 ... (2022-03-22)
A.Buchanan: Hi Mario, thanks I agree (2022-03-22)
A.Buchanan: How about sBc6 to fix it? (2022-03-22, hidden)
A.Buchanan: How about +sBc6 to fix it? (2022-03-22, hidden) more ...
comment

SCHRECKE: Konnte zuletzt nicht auch 0. ... Kg7-h6 geschehen sein? (2021-02-15)
A.Buchanan: I think the diagram must be wrong. Not only does the retro logic fail, but also there is a dual 2. Df3+,Dg4+. It's not in WinChloe. (2021-02-15)
Mario Richter: An additional white Pawn e7 would justify the epkey, but not prevent the mentioned dual. (2021-02-16)
VL: We might shift wPf2 to f3 against the dual. Or exchange Bf6 with Pf2 and add wP on f3 (instead of e7)... (2021-02-16)
A.Buchanan: Shifting wBf2 to f3 leads to short cook solutions 1.Dh2. However I think 8/5p1p/4NB1k/3PPKpP/4P3/8/5PP1/2Q5 fixes the retro issue, and it's C+ by Popeye, with still ideal mate! The composer Tomislav Petrovic is still alive but would be very old (90). Is he still active? (2021-02-17)
A.Buchanan: By the way, if you query for k='foo', that will always return uses of the keyword "foobar". It's just checking for the prefix. If you omit k='foo', then that will also block "foobar". Therefore if we tag with "en passant as key", then we should not tag with "en passant" too. Both keywords should exist, and over time we may refine uses of "en passant" to be more precise.
Because of the sensible way prefixes are handled in PDB, this is a different situation from tagging a problem as both "ceriani-frolkin" and "prentos", which we should do.
On this general subject, when counting instances within a single problem, we should always use Gerd's excellent mechanism of parameters "foo:1" etc, rather than having separate keywords "foo 1", "foo 2" etc. Otherwise the set of keywords becomes unmanageable. (2021-02-17)
A.Buchanan: wDc1 fixes the dual but the retro soundness is another matter. Current position has ideal mate, which it would be nice to retain (2021-02-17, hidden)
A.Buchanan: By the way, if you query for k='foo', that will always return uses of the keyword "foobar". It's just checking for the prefix. If you omit k='foo', then that will also block "foobar". Therefore if we are having "en passant as key" as a keyword, then we should have "en passant" as a keyword too. Because of the sensible way prefixes are handled in PDB, this is a different situation from having both "ceriani-frolkin" and "prentos" as keywords. (2021-02-17, hidden)
A.Buchanan: By the way, if you query for k='foo', that will always return uses of the keyword "foobar". It's just checking for the prefix. If you omit k='foo', then that will also block "foobar". Therefore if we are having "en passant as key" as a keyword, then we should not have "en passant" as a keyword too. Because of the sensible way prefixes are handled in PDB, this is a different situation from having both "ceriani-frolkin" and "prentos" as keywords. (2021-02-17, hidden)
A.Buchanan: By the way, if you query for k='foo', that will always return uses of the keyword "foobar". It's just checking for the prefix. If you omit k='foo', then that will also block "foobar". Therefore if we tag with "en passant as key", then we should not tag with "en passant" too. Both keywords should exist, and over time we may refine uses of "en passant" to be more precise. Because of the sensible way prefixes are handled in PDB, this is a different situation from having both "ceriani-frolkin" and "prentos" as keywords.
By the way, we should always use parameters "foo:1" etc when counting instances within a single problem, NEVER have separate keywords "foo 1", "foo 2" etc, or the set of keywords become unmanageable. (2021-02-17, hidden) more ...
comment

Henrik Juel: C+ Popeye 4.61 (2021-04-14, hidden)
A.Buchanan: I think this vanilla AP-Petrovic works well based on RS. I don't see PRA is applicable at all. (2022-03-24, hidden) more ...
comment

A.Buchanan: A small group of similar problems forces some short Round Trip across past and forward play. The last move is not uniquely defined (it may be a capture) and is not mentioned in the stipulation. However the genre includes retro. This means that an aesthetic criterion (e.g. existence of Round Trip) can define whether a composition is subject to 50 move rule or dead position rule! (2021-10-31)
comment

Yuri Bilokin: Compare P1189828 (2024-04-03)
comment

SCHRECKE: C+, popeye 4.87 (2022-10-22, hidden) more ...
comment

Anton Baumann: C+ Gustav 4.2a (2021-11-29, hidden) more ...
comment

A.Buchanan: This is definitely not h#0.5 rather it's h#0.25 or #0.5.
This is retro joke not fairy, imho (2022-03-03)
comment