Mate in two (634)

You're welcome sameez1!  But it was not about this puzzle which was great!

I did not think there was much of a chance of you seeing my comment on the daily puzzle thread..In defense of us weak players though I would like to add that the daily puzzle has been using Queen capture as solved,.

Btw, I think the daily puzzles are great since the current editot has taken over. There is easy, medium, hard and very hard in a good mix. And they are all correct unlike before!

And that I said the puzzle is "easy" for weak players does not imply that anyone who finds it easy is a weak player. I am sure Magnus will find it easy as well.

Note on the joke rules. The example of "not having a dead rule" in my last post might be confusing. Any rule could be a joke rule in one context and not in another but this one is likely outside the thin atmospheric layer surrounding the planets of our systems. Sort of boring. Joke rules are not part of the system but they should pleasantly rhyme with the system. Would have been more accurate to state that any undisclosed rule is a potential joke rule if its implementation is clear and natural. Also, a potential joke rule only becomes an applicable joke rule in a given problem and is therefore dependent on 3 objects: {system, potential joke rule and given problem}. I see no difference between fairy and orthodox systems in this respect. I bet that Chess960 will one day have its own joke rules when people get fed up with just "obeying the rules".

Hi Ari: on Chess960. What is your castling convention for Chess960?

The conventions which address proof game doubt are the closing pieces in the information system for compostitions. The other sources are rules, stipulations and general context.

I suggest - and I am probably not the first one to suggest that - that the primary means for providing PAS information in chess960 would be in the form of a 1st rank FEN mask in the stipulation as specified by the author. Examples:

[BBRQNKRN] is full FEN information, [xNxKxRxx] is masked information where the "x" is unknown, or just [x} indicating that no 1st rank information is available. Without specified FEN it's the orthodox PAS. This way integrates orthodox and chess960 rule sets.

The castling and other conventions operate as an extension to the known rules and the masked FEN information. The basic chess960 castling convention is the same as in orthodox chess: any castling is permitted unless disprovable on the basis of the available FEN information.

The default interaction with all other conventions would be advisably based on RS logic: "come first, served first" amongst preferential assumptions (castling permission, no e.p., no 50M or 3R draw). If you want it, PRA logic is also the same as in orthodox chess but with unknown rook posittions there is an enormous amount of parts. AP is also as usual. There is no reason to adapt the big logics for chess960 as they are generic concepts and can handle many different rule systems.

I pretty much agree with all of this.

The point of divergence is that the mass of Chess960 problems today do not include a mask. The composer might say: well the default is that those pieces on the back ranks now in the diagram began on those squares, unless you can prove otherwise. No one for a Chess960 problem would say that the default is the orthodox PAS. They would try to extend the Constitutional Law which applies today to whether K&R have moved (very reasonable) to a much bigger ask: that pieces began on the squares they occupy now. And would this apply to just K & R, or all pieces on the back rank?

That's the real issue here

I didn't intend to say that the default for a chess960 problem is the PAS. It would only be the PAS if you chose not to separate the orthodox type from the chess960 type in the stipulation. So it could be used to unambiguously unify orthodox and chess960 diagrams. Btw, any FEN-string - like simply [x] - could be used to identify a chess960 composition rather than stating "chess960". Talk about efficiency!

I have thought of the possibility that authors would want to make the assumption of the diagram 1st rank but I didn't know that option is already in use. It is very easy to integrate this method in the mask-system because the diagram acts as a mask in exactly the same way as the masked FEN. You can enter it in the stipulation as [d].

The problem with diagram masks is as you indicated that there is no obvious justification for chosing the 1st rank over the 8th rank or vice versa. And they may be in conflict. I don't think it's a good system, just a lazy system. But it fits in the generalized masking approach if you want it.

Addendum: The "real issue" with the diagram array is much bigger than just the castling units. How you fill in all the PAS pieces also affects the retro-analysis of castling and e.p. right and possibly 3rep and 50M situations. The correct way to address them is not by designing ways to interpret the diagram but to use the retro-logics properly. For instance, instead of worrying about whether or not a rook is on its initial square, you consult the convention which tells you when you are permitted to castle. In almost all cases this resolves the question you need answers to in relation to castling "right".

Note that most of the remaining choice issues in identifying diagram PAS units are simply undecidable through lack of information. My original FEN proposal opens the door to specifying precisely which units are in PAS positions, which are not and for which it is unknown by design. The toolkit is complete, why make things anymore complicated?

I get your mask mechanism and I am sure it’s quite general but I don’t think people want to compose problems like that. It’s not about the efficiency it’s about the naturalness of the problem stipulation. “People” will want to extend the existing castling convention to Chess960. Is the starting location a Constitutional or Police Court matter? I.e. is it a right by default or a wrong by default?

Edit: I now doubt I got the definition of the Constitutional Law right. Lost the article long ago and can't find a reference. If I am wrong, the post needs to be rewritten!

Edit2: Ah, I figured it out. Read the next post!

I know another topic where people didn't like masks but use them now anyway

The arguments against masked FENs are weak since they focus on castling which is easy to specify. When you consider the complete retro-analytical spectrum, it is obvious that you may need to know more about the PAS than just the king and rook positions. It is weird to assume that retrograde composers in particular will generally want to synchronize the 1st rank of their diagrams with the PAS of the chess960 setup. Nobody says you must use FENs in all problems but they are a powerful weapon in the arsenal.

In my approach no FEN supplement is needed for castling rights. By the traditional RS-concept (which is much broader than castling as demonstrated by Plaksin and myself) castling permission exists in all chess960 diagrams if not disprovable in all available chess960 setups. RS-logic is superior over constitutional rights in dealing with these issues. The latter is only useful in fairies where retro-analysis is unreal and needs to be replaced by something static. It's a surrender flag. We can't get it done, let's postulate a fixed state. I don't recall the precise Police Court definition but it might very well be similar to RS-logic. Since some of the abominable problems were flawed, It was hard to figure out everything the designers intended.

Examples of victims to the constitutional castling law if it applied to the orthodox domain: PRA-logic, AP-logic, mutual exclusive castlings, Plaksins castling+50M problems, most of my retro-problems and many basic RS-problems. The devastation would be immense. Next think of all the compositions that therefore can't be made in chess960!

I think there is still a communication gap. The basic idea of Constitution & Police Court, due to Ronald Turnbull (who I met physically for the first time in October) is simply to say for any kind of conditional move: is it allowed or disallowed by default. So you can't get away from the responsibility for assigning any type of conditional move to one camp or the other. You've just said that castling is Constitutional in Chess960 over all possible PAS. That's the answer to the question I posed earlier. That's fine: I just wanted to know which side you come down. There are no "victims" in the orthodox domain: to say castling is constitutional while e.p. is police court is simply to describe the actual situation. Then when one encounters other conditional moves (e.g. in fairy domain) a similar choice can be made

Addendum: I found some notes on MatPlus and now I know again how everyone is thinking on this funny planet. There is no need to make a distinction between "Police Court" and "Constitution". The illusion that they are different comes from the associated actions. One emphasizes an action, the other one an inaction. However, fundamentally all that counts is that both actions and inactions are state-related. The possible states for both are the same: you can either prove that castling, e.p. or anything is valid in all proof games, in some proof games or in no proof games. Every property gets an arbitrary handle which says whether or not an associated action is permitted for a certain state. All these handles are property based, not global. There is no such thing as a constitutional or police court view of Chess960. There are only individual retro properties with an action handle. The really big global logics (RS, PRA, AP) may override the individual property handles and are therefore way more important. For instance, the e.p. handle commonly requires full proof, but not when the logic is AP or PRA. Even in RS-logic the e.p. move may still be valid when its right must be elevated due to competing e.p. moves.

Though the distinction between police court and constitutional is sort of insignificant, one might still insist on using them where they cause no trouble. One issue is then of the greatest importance. Is the state (e.g. the castling state) only definitive after the constitutional move is executed or could it be lost when not executed by an another event in the solution? The answer to this question is absolutely critical for rating the constitutional law especially when evaluating diagram states. If the castling "right" cannot be lost then it is a fixed state, otherwise it is adaptive. Fixed states destroy retro compostions, adaptive states do not. Can't answer that question and therefore can't endorse or reject constitutions or police courts  

OK, we are cross posting again. Your answer provides clarity on some some issues not on all.

I am aware there are no victims in the orthodox domain because the constitutional/police court laws do not exist there. I am not sure there wouldn't be if they existed. See the last issue in my prev post.

And yes, there is a communication gap due to the paradigmatic chasm!

And yep, I understand Ronald Turnbulls viewpoint but if that is the full thing, it lacks state appreciation. For instance, is he OK with the fact that different variations in a solution have different histories with different property states? Again, see the last issue in my previous post.

And a note: It is OK to fix (a part of) the state of the PAS position by a diagram or a FEN mask. It is just not OK to fix the idea that K and R did not move during any of the proof games.

Constitutional / police court is just a binary classification of existing stuff. They are not laws so can't break anything. Castling and e.p. conventions do have different state for different RS variations. In PRA it's all retro logic is all pre-processed. In SPRA the same except en passant now becomes constitutional.

In what sense is it "not ok" to include in a stipulation that K or R did not move? It may not be very aesthetic, but it defines a subset of the states just as well as FEN does. The fact that you feel the need to exclude stuff like this makes me think your theory is more fragile than universal. There is a whiff of the "No true Scotsman" fallacy here.

The inability of your theory to absorb changes in rules and conventions over a long term is also a red flag for me. Eppur si muove!

Beginning sounds great, the end is another paradigmatic miscommunication.

I don't worry about binary classifications as long as they do not exert decisive force. I never use SPRA and I always forget what it means. Can you tell me in short?

The miscommunication is about generic versus instance based criteria. Everything is OK in an instance (a particular problem) but many thing are not OK in a mandatory umbrella rule. For instance, if everyone uses the diagram FEN mask for lack of alternatives and its use would imply that K and R (say on the white side) have not moved then you would have taken a free variable out of the equation. Then you can no longer prove that castling is illegal while the position is legal. And all the other things I mentioned. That is why I was curious about the precise definitions of "constitutional" and "police court". Note that the choice of the term "constitutional" radiates an image of "unnegiotiability" while such is apparently (hopefully) not intended. I am a defender of freedom, not an attacker!

If you think my theory cannot absorb changes in rules and conventions you have understood none of it. It is a meta-system to an immense choice of rules and conventions and I don't think anyone is currently near its boundaries. It does however give some protection against inconsistent, unsound and incomplete concepts.

To give an example in Turnbulls domain: Turnbull was very much focused on what you can do with units which possibly carry a retro-property like being-fuddled on a particular move. What he missed is that the complete set of conventions should plug all information holes in the system not just the one you are focusing on. The fuddled men (and many other fairy rules) require additional king move conventions. King moves become subject to proof game doubt when they move into the range of a possibly fuddled man. There are 2 obvious approaches, (a) the king cannot move there (b) the king can move there and thus declare the man fuddled. Both have consequences, for instance, what will you do with "(a)" if your king appears to be stalemated by units one of which must be fuddled? Reminiscent of the mutually exclusive e.p. moves riddle. Turnbull is a supposedly intelligent man. Why didn't he think of this at the time (may be later?) though it affected one of his problems - while it was the very first thing I looked at? The answer is that his attention was focused of Constitutional and Police Court rules for his new retro-property and not on the holistic operation of his complete system. And I believe this is true for the whole community. Everyone rushes in to produce new compostions, invents some rules/conventions to get started, and only reflects when troubles come in plain sight. The purpose of my (methodological) paradigm is to see the troubles coming from a 1000 miles and to forestall them.

Note: Just think of it. My terms for constitutional and police court are dominant and recessive though I hardly ever use them. When you think of the biological inheritance methaphor where the different parents constitute the different proof games and the offspring is the composition diagram, you will see these terms are perfect. Why does nobody read my suggestions? (rethorical)

OK quick response. Dominant/recessive metaphor works for me once you communicate that someone can have n parents. "Constitutional" has no more baggage than "dominant" here - just a classification. SPRA is like PRA except e.p. right is dominant. I'm sure it was you told me the other day that your system was fixed and so couldn't handle change, but I can't find the mention. But you've skipped the point about Smullyan-style stipulation for chess960 being "not ok". Turnbull seems pretty bright to me, but he is definitely not a formalist. He is very focused on composition, and doesn't seem to mind a few unresolved paradoxes lying around. To me, the two of you make an interesting comparison of two very different very creative minds. He has given up chess composing because he decided he could have more readership by writing hill-walking books.

I see what you mean about free variable. Yes there are two layers in castling rights for Chess960: initial locations and whether K/R have budged. Can't conflate them. This is exactly the point I was trying to get at before. Whether K/R have budged is normally constitutional, but it seems like taking a liberty to assume that the diagram locations for pieces were the original ones.

I don't recall all the things I wrote to you (there was a lot) but if I wrote my system was fixed then I would have been intended as a meta-system. Which means it can handle infinitely many different rules and conventions but also that there are limitations. If something needs to be changed on the meta-level I might need to revise the paradigm. Never know what the consequences of a tiny change are. Note for instance that I can handle AP-logic only with considerable effort since it can't ordinarily be done with FOL (first order logic). Mostly I rely on relational database tables driven by FOL as prescribed by Codd.

I have scrolled back some posts but I can't find anything about Smullyan-style stipulations. I presume you refer to fixing certain rights like 0-0 and/or 0-0-0. Can't imagine I would oppose that as all stipulations occur in particular problems and I do not oppose anything in particular problems. There is only the general notion of quality. Problemists will perceive any and all extra information as a weak spot in a composition. That would also be the case with extra masked FEN information. But in the latter case it is obvious that you need some way to reduce the massive load of chess960 uncertainties either by a diagram mask, a default mask or whatever other way. Which reminds me of a nice type of chess960 challenge: "SPG in 22 moves; what was the PAS?" Need no mask of any kind. Need not even an unambiguous proof game.

It is OK to let the diagram locations be the original ones provided you can resolve all ambiguities and contradictions. I will add some diagrams at the end of this post. Please tell me which pieces are in the original locations! There is no duty to treat PAS variables in the same way as the proof game variables. It's a separate phase and therefore a separate choice. It all depends on your system model. If you see the PAS setup as the first move, then you can treat it in line with the other moves. But you can also handle it as a distinct layer with its own conventions.

Great idea to write hill-walking books. Unfortunately no hills in my area. I'll try below-sea-level-cycling-books!

Chess960 diagrams with diagram mask convention. Which pieces are in the original locations - irrespective of whether they moved or not?