"...then Answer A is retroactively no longer a correct answer." How so? It's not wrong unless it doesn't solve the problem posed in the question; it's just no longer the optimal one. Someone can still solve the puzzle later and try to optimize their answer; they just might not come up with the most optimized answer. I don't see a non-optimal answer as any different than a partially solved crossword. The answerer got partway there, just not all the way. (I don't know what's best for this site, but that point just seems wrong to me.)

@noedne Are you sure? Less optimal answers are often important steps in reaching the optimal one.

@jpmc26 not all open-ended questions are optimization problems. Optimization problems have a provably optimal solution, while most open-ended questions have an arbitrary set of criteria to determine the “best” solution. Optimization is OK; open-ended is not.

@PiIsNot3 The definition states that, "answers are ranked by some sort of rule." This rule is the "function" to be optimized, even if it's not presented as a formal mathematical function. By definition, it would seem that all open ended questions must be optimization problems of some kind, though they may not be provably optimized. Regardless, my point is that answers are still not invalidated by a new, higher ranked (according to the rule) answer being given. They're only invalid if they fail to meet the other constraints.

@jpmc26 The big issue with that is no single rule or even set of rules is enough to guarantee there be a single correct answer. For “longest word/sentence” type problems, the answer changes depending on what dictionaries you use and which words you allow. For “formation of numbers” problems, you have to define which operations to allow, what constructs are valid/invalid, etc. etc. Most questions tagged open-ended don’t have that level of rigor and instead devolve into a game of “what loopholes can I exploit that give a technically correct answer.”

In contrast, optimization problems are often grounded in math and logic, two subjects that have very clear-cut rules that don’t allow many loopholes. So that’s the difference between optimization and open-ended (or at least, what my understanding of the difference between the two is).

@PiIsNot3 The core of your point seems to be: "Most questions tagged open-ended don’t have that level of rigor and instead devolve into a game of 'what loopholes can I exploit that give a technically correct answer.'" That's a fine objection to the tag, but it has nothing to do with making one answer invalid by posting another one. As I said, I don't have any answers for whether the tag is bad or not; I just found the first argument to be very flawed, which hurts the persuasiveness of this answer.

@jpmc26 It’s an unwritten rule on PSE that all puzzles must have a justifiably best solution. Allowing accepted answers to be beaten by “better” ones goes against this rule, as Deusovi alludes to. Perhaps that section of the post should be moved to the top to make the point clearer.

@PiIsNot3 So most match stick puzzles should be removed? This word puzzle should be closed? It seems to me like many popular, accepted puzzles have no provable single answer.

The match stick puzzle is personally very iffy to me, because the rules as they stand allow many possibilities. The word puzzle does have a clear end goal at least, but again the rules allow for a broad range of answers.

I think if you insist on every puzzle having a best answer that can be proven (at least informally) to be the best, you're going to cut out a large number of puzzles that really don't pose any practical problems and seem to be quite well liked by your user base.

And whether or not a puzzle is popular or not should have no bearing on the quality of the puzzle. There are plenty of examples of well-constructed puzzles on PSE that barely get any views, while riddles like this one get a free ride to HNQ

Or an even better example: this one

Puzzling.SE is primarily a site where people have fun doing puzzles. A puzzle's popularity is a reflection of how much people are enjoying it. While many SE sites are right to cut popularity out of their on-topic consideration, I'd argue it's a very good consideration for Puzzling.SE.

That doesn't mean allowing all popular puzzles just because they're popular, but it does mean it would be counterproductive to ban entire classes of puzzles just because of a technicality like "no provably best answer."

Of course, I wouldn’t know how open-ended questions are actually affecting the site (I’ve only been here for a month, after all). But, like Deusovi mentioned in the post, if the site were overpopulated with those kinds of questions, I wouldn’t imagine it drawing the interest of many people

At least, the kinds of people that make the site what it is now

I agree there are convincing reasons to get rid of open-ended. There seem to be pervasive quality issues in it. I just don't think the idea that there must be a provably correct single answer is one of them, as it would imply removing vast swaths of many other categories as well. Throwing the baby out with the bathwater, so to speak.

I think it's fine to say that an answer must just be "correct" in the sense that it meets all requirements the puzzle specifies, and then allow for multiple of those. That doesn't seem to be particularly problematic.

I think rather that insisting on a single correct answer is an outgrowth of the fact that the platform allows for accepting a single answer, but this feature of SE doesn't really have much relevance to Puzzling.SE's goals of offering high quality puzzles for people to enjoy.

I guess I do agree with you on that front. There are some examples of riddles on the site that are well-received that have more than a few alternative solutions. However, the caveat there is that the riddle doesn’t allow for overly broad answer; you still have to have your answer satisfy all of the clues.

The riddle I linked earlier was closed for being too broad - there were just too many possibilities that fit the clues as stated

Sure. And if a puzzle is so poorly formulated that half the answers consist of exploiting silly loopholes the asker failed to close, that's probably not a good puzzle.

Same with riddles where there's so little detail that any number of things can answer it.

From my limited experience here on this site, the most well-received puzzles tend to be those that have clear-cut answers that fit all the clues and hints as originally posed

The best riddles seem to be the ones where there's sort of buried, subtle little hints that once you see them in an answer, it's obvious that's what the asker was trying to lead you toward. So I do get what you're saying, but I just don't think open-ended is the same sort of beast in that regard.

There are a few examples of “good” open-ended questions on the site. This one’s probably the best of them all. But even that one suffers the same issue of there being the possibility of a new better solution cropping up, thus rendering all the previous answer invalid.

But again, an answer isn't invalid just because it doesn't have the best score.

It's only invalid if it goes outside the boundaries of the used digits or operations.

How about the other point brought up in the post? That open-ended questions aren’t really puzzles but rather are games where you have to beat others’ answer rather than try to find a specific solution

I don't know if games are on or off-topic on Puzzling.SE. Codegolf.SE is basically one giant game, and they seem to be managing it without too much trouble. So having them be on topic is probably feasible, if a set of criteria for evaluating them can be formulated.

Hmm, that’s an issue that’s probably best left to the mods and more experienced users

The pi post seems to be in the spirit of Puzzling.SE, whereby some challenging set of constraints (if you'll permit me to think of clues in riddles as "constraints") is given and users are encouraged to find a solution that satisfies them.

If you were going to argue against the pi post, I'd say probably the best argument is that it has a gaping hole in its formulation. In terms of open-ended in general, preventing these kinds of problems is extremely difficult, which gets back to your point about loopholes.

On the other hand, it can be fun itself to find and exploit this kind of loophole when it's not overdone and there aren't many of them. So I don't know.

Perhaps you should write your own reply to the meta post? I see you have a lot of good arguments for why the open-ended tag should remain on the site :)

I'm not really arguing for that. I just think you have to be a lot more careful about how you argue against its usage. Settling on a particular argument leads to precedent, and if those arguments undermine a lot of good posts, then you start making the site less fun. Which is wholly against the spirit of the site.

Again, I’m just a newbie here, so I don’t have much insider information on what’s considered on-topic in PSE or the overall consensus opinion on what PSE should be for. I think it’s best to take this conversation elsewhere to users with more knowledge than I do

I'm even more a newbie than you. I've never posted a question or answer here. XD

I just like to think moderation problems through. lol.

Welp, you got me there haha :)

Hopefully Deus will see this chat log and make adjustments as necessary to the post

BTW I did some digging around in Meta and found these posts which relate to the topic at hand

Hey! I'll try to give my thoughts on the discussion here:

@jpmc26 The question asks for the optimal answer. So if it's not the optimal answer, by definition it cannot solve the puzzle. And non-optimal answers generally can't be added to to give an optimal answer. They're not "part of the way there", they're just "a separate way of doing it that isn't optimal". You would have to change the answer significantly to make it into the optimal answer.

@jpmc26 Any matchstick puzzle of the form "make the biggest number you can by adding/moving/deleting n matchsticks" should be removed, yes. (I'm perfectly fine with "make [this geometric figure] by [doing this thing with matchsticks]", though. And in my experience, most matchstick questions are the latter type.)

To elaborate on the first point: The question is "accomplish this task in the optimal way [that PSE members can find]". I'll make a distinction here between a strategy, which is a method of accomplishing the task, and the solution, which is the best strategy of accomplishing the task.

Every answer posted contains a strategy that the poster believes to be the solution. And it is the solution, but only temporarily: only until it's beaten by another answer.

This type of thing is fine on sites like Code Golf, where the site is intentionally structured around these: every post is not a puzzle, but a game. (Or a "challenge", if you like.) There is no one optimal solution, and in fact the official policy there is that all languages are competing separately (so there can be a 'winner' for Python, a 'winner' for Haskell, a 'winner' for Befunge...)

That is a direction that we could take the site if we so chose: after all, it worked for Code Golf. But that seems to me to be disregarding the fact that puzzles should have a definite solution. (In my eyes, having a solution is a necessary quality to be a "puzzle" in the first place.)

(And as a side note, I'd like to mention that I would likely no longer be interested in PSE if we did take that direction, and I believe I've seen at least one other prominent member who echoed that sentiment.)

...Anyway, my point was that every answer has a strategy that may or may not be the solution.

The newest strategy is constantly in a state of limbo until either people lose interest (in which case it is considered the solution), or it is outdone by another strategy (in which case it is definitely invalid: it is no longer even possibly the solution, and therefore is not the optimal strategy).

A few other things I wanted to respond to:

@jpmc26 "users are encouraged to find a solution that satisfies them." No, users are encouraged to find the solution that satisfies them. This is an important distinction. If a riddle has multiple equally valid solutions, we close it as too broad.

@jpmc26 Yes, a bit. So far, those questions have been an anomaly, though - an occasional exception to the rules. I think there's no reason to make that exception, and wouldn't be happy if that occasional exception turned into a much larger part of the site.

Also, you seem to be treating "single correct answer" as a "technicality", or something invented for my meta answer. But this is the policy we use for all other puzzle questions: for example, if a riddle has multiple equally valid answers, it's closed as too broad. This is one of the core principles we have to determine whether a question is appropriate for PSE, but we seem to be ignoring it for the case of the open-ended tag.

Good discussion @Deusovi @jpmc26 @PiIsNot3

@Deusovi Has there been explicit prior Meta discussion about that? The only thing I can find that's in that vein is an old answer of yours, but the context of the discussion seems to have been limited to riddles (perhaps accidentally, but that seems to be the case nonetheless). Only one answer here arguably covers single correct answer, and even it doesn't say so directly.

It seems like discussing that principle and getting community input would be a good thing to do. Some research into what kinds of puzzles it would impact and how prevalent they are on the site would be worthwhile, as well.

It just seems unlikely to me that most people who enjoy doing puzzles are particularly turned off by problems that intentionally have multiple possibilities for solutions. I certainly understand the problem where there are multiple solutions just because the problem isn't specific enough, which reduces them to a pointless guessing game, but there's another kind of problem where the constraints are clear and the freedom within them is intentional.

A few good examples were linked here. These seem to be liked when done well, and conflating the two seems counterproductive.