When writing riddles, I often encounter an issue that I've decided to call the "Riddle-maker's Dilemma." (Forgive me if there's already a name for this.)

In short, the line of thinking goes like this: I want to make my riddle difficult, so I need to make the clues more abstract and metaphorical. However, making the clues more abstract also means that I'm increasing the number of almost-answers, that is, answers that kind of fit the clues but aren't what I'm thinking of.

I firmly believe that the answer to a riddle should fit like a key in a lock; when you hear the answer you should think "Oh, I should have thought of that," not "C'mon, my answer is just as good." The top-voted riddle on this site is an excellent example of the good kind of difficult riddle I'm talking about.

How can I make my riddles difficult without relying too heavily on abstraction?

Or, is this a false dilemma and not really an issue?


4 Answers 4


I looked up the 10 highest-voted puzzles here that are tagged [riddle]. Warning: slight spoilers for those puzzles lie below.

I draw the following conclusions:

  • Many of the best-liked puzzles here that are tagged as riddles have important non-riddle elements. Making a puzzle more than a riddle may help.
  • The commonest way to make there be a single clearly-best answer is to make the puzzle appeal both to meanings (which may, indeed, be kinda vague and abstract) and also to something else, most commonly wordplay. That means that for a solution to be as good as the intended one, it has to match both the semantics of the clues and also the largely-independent wordplay elements, which is very unlikely to happen by coincidence. (This is the same principle that makes cryptic crosswords work.)
  • Sometimes, what makes a puzzle work is something specific and one-off that can't easily be fitted into any sort of general theory of how to make a successful riddle.
  • $\begingroup$ The second one you've listed does have a sort of riddle at step 3 in the solution. $\endgroup$ Jul 5, 2018 at 19:09

There's a couple things to this question and I think we can treat them separately from each other. Your principle question, really, is "I want to make my riddle difficult" - the rest of it is specific questioning about a specific approach you are considering for achieving that. So let's look at that first question first.

Difficulty is, well, difficult. It's a subject that comes up with regularity, most recently here: How to make a riddle tricky without too many clues to make it easy but enough to avoid "too broad"?
The accepted answer to that question has a lot of thoughts on puzzle difficulty and how to make a puzzle not merely more difficult, but better. Rather than reiterate everything that was said there I'll just point you to that answer.

For your second question, and the crux of your "dilemma" - how do I make more abstract clues, so my riddle is harder, without making my riddle too broad?

A couple of canned thoughts, that I mention with some regularity in comments when people have gotten this wrong:

  • It's not sufficient that your intended answer fits all parts of the puzzle. If many answers could fit, then the puzzle is under-specified. A well-crafted puzzle will give enough information to rule in the intended solution while ruling out everything else.

  • [Some] riddles often allow good interpretations and answers that were not intended, as is happening here. If there is some reason (beyond "that wasn't what I was thinking of") why the existing answers do not solve this riddle, it's not apparent at all what part of the riddle actually invalidates those responses. Because it seems they should be at least as valid as any other answer you might have in mind, this may be "too broad"—you may need to update the riddle to make sure invalid responses are demonstrably invalid.

The earlier answer I referenced, from point 4 (When is a riddle really "too broad"?) onward, delves deeper here to what you need to do, and why, to make sure you don't end up with something too underdefined. (See also the other answer to the same question, which discusses avoiding making it overdefined as well, as that tends to lower the difficulty.)

But there's little guidance here on how to do that, as there are a number of good ways to do so and how a specific puzzle crafter chooses to achieve the goals are likely to differ greatly from someone else's choices, depending on their skill in using different puzzle elements. @Gareth McCaughan's answer here offers some excellent observations for how to make your riddle be more than just a riddle, so there are different paths a solver can work at to move toward a solution even in the face of more abstract clues; that's certainly a great way to go. Other riddles often end up adding hints that give additional clues toward the solution as time goes on, but as noted elsewhere these tend to feel like clumsy patches to the puzzle's difficulty; ideally, offering enough clues that converge to the same solution from different angles as part of the initial puzzle is the best way to go.

One suggestion that I can't recommend enough: ask a few people to try to solve your puzzle before you post it here! That will give you some great feedback on what people find immediately obvious, what people have to think about, and what puzzle elements they never figure out. Such feedback will help you avoid making things either too trivial, or impossibly inaccessible to anyone not reading your mind. Once you've tried this with a few puzzles and start to get a feel for what works well and what doesn't, it becomes somewhat easier to guess how a new puzzle will fare, but until you've gotten a feel for it, test solvers are very useful!

  • $\begingroup$ Discussion of underspecified riddles reminds of of the brilliantly humorous dialogue between Merlin of Chaos and the Sphinx in Roger Zelazny's The Trumps of Doom. If you google "trumps of doom frog in a cuisinart" you'll find it, although I don't like to link to the copyright-violating results. :) $\endgroup$
    – Wildcard
    Jul 5, 2018 at 23:14

You may want to check out some of the riddles by Hugh Meyers.

He is, in my opinion, the best riddle-setter on this site. Most of his riddles have a "gimmick". In addition to the description given in the riddle, there is some sort of wordplay or hidden message that unambiguously confirms that you have the right answer once you've found it.


The dilemma definitely exists, and there are several ways to cope with it. These are my favourites:

Plan 1: "The checksum approach." Go abstract, make fun clues that can yield ambiguous results, but then, have a checksum to verify the answer: eg. a couple of wordplays that cannot really be solved without knowing the answer already.

Plan 2: "The literal approach." Keep things literal and/or direct. If you take poetic license on even one clue, you open up all of the clues for similarly vague interpretation. As long as all of the intended answer's explanations are more or less direct, any overly inventive and/or lateral-thinking clues are easy(ish) to discount.

Plan 3: "The opposite approach." I've recently been experimenting on intentionally making my puzzles easier rather than more difficult: I realised that the reason I wanted to make my puzzles more difficult was that I felt that the added effort increased the satisfaction gained from figuring out the solution. Working on that basis, I then tried to create puzzles and riddles that are simpler to solve, but will offer some kind of an additional silly reward as a part of the process, to account for the relative lack of effort required. If that sounds too abstract, you can take a look at my recent questions: for example, in this wordplay riddle you have to first solve the riddle, and then you get to hunt for the remaining clues afterwards.

There are other approaches, of course, these are just the ones that are the most pleasing to my (admittedly more literal-minded than many) taste.

  • 1
    $\begingroup$ I fundamentally disagree with both "Plan 1" and "Plan 2". Plan 1 is basically "make a crappy riddle, then let people keep guessing till they find the answer you were thinking of." Plan 2 is "let's play 20 questions". Neither make for a good riddle. $\endgroup$ Jul 5, 2018 at 18:18
  • 1
    $\begingroup$ @GentlePurpleRain your comment comes across as unnecessarily hostile, are you having a bad day, maybe? My answer is to someone using the phrase "riddle should fit like a key in a lock", so I think it's quite reasonable to assume that they won't immediately go ahead and make the worst imaginable puzzle while still adhering to the letter of the plans. $\endgroup$
    – Bass
    Jul 5, 2018 at 18:42
  • 1
    $\begingroup$ Sorry if I sounded hostile. I probably could have worded that better. I do disagree with both of those Plans as methods for constructing a good riddle, but I could have said it more nicely. The point I was trying to make (not very well) is that the riddle's solution should be obvious (once found) on its own without requiring verification (Plan 1), and that a riddle necessarily requires some metaphor or obscure language, or else it is not a riddle and is simply a list of characteristics of an object/idea. $\endgroup$ Jul 5, 2018 at 18:46
  • 1
    $\begingroup$ @GentlePurpleRain Thanks for elaborating on the point! I still don't see how those things are excluded by the plans though. If there is one answer that Really Clicks, and other, plausible ones that basically hit the clues, but don't really click too much, it's simply common courtesy to explicitly exclude the unintended ones by giving a wordplay that definitely rules them out before someone posts them as answers. I may have some opinions on the necessity of obscurity, but this comment is too small to contain them. In any case, I am not talking about property lists when I recommend direct clues. $\endgroup$
    – Bass
    Jul 5, 2018 at 19:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.