I don't believe this to be a meta question. It is about puzzle creation. Although I mention two puzzles that I have already set, I am not complaining about down-votes, I'm trying to see what aspect of the design (if any) separates them in popularity.


I have submitted two puzzles that are based on Google ngrams. The first got lots of upvotes and the second got lots of downvotes.

Assassinate Shakespeare!

ngram puzzle No 2 (and they are getting tougher!)

I have an idea for a third but I want it to be successful.


A. Can anyone explain the different attitudes towards the two questions in terms of up- or down-voting?

I have a couple of theories

  1. Maybe one ngram puzzle was enough and members are now bored with the idea.

Note that the questions don't have an identical format. The first question relied on a cryptic clue whereas the second require a deductive leap. You had to notice the peaks of the curves and then work out the significance of those peaks.

  1. Maybe the second puzzle was too considered too difficult by some people and they down-voted for that reason.

Clearly it was possible to solve because someone did. But would people really down-vote a puzzle just because they thought it was too hard?

B. If I post a third it will still use ngrams but the solution strategy will be different yet again. So, how can I design it so as to ward off down-voters this time? Just what were they objecting to and what was so good about the first?

  • 18
    $\begingroup$ I fail to see how a discussion of 2 previous puzzles is not meta.. Especially since you don't discuss creation as you say it, just mention it. $\endgroup$
    – DrunkWolf
    Oct 14, 2015 at 4:09
  • 3
    $\begingroup$ @DrunkWolf - Thanks for your comment. I gave some considerable thought to this but came to the conclusion that here was the right place. It is difficult to discuss puzzle creation without examples. There are precedents, e.g. goo.gl/x9gwTK ---> Essentially my question is, "How do you make a puzzle popular?" My first was popular, my second wasn't. To me they should have been equally so. The answers below, tackle this question and in fact provide some very useful tips for my next puzzle. I'll soon test them out and see if they work. $\endgroup$ Oct 14, 2015 at 7:55
  • 1
    $\begingroup$ Your first puzzle was named : Assassinate Shakespeare!...... Enough said...... $\endgroup$
    – Pieter B
    Oct 14, 2015 at 8:39
  • 2
    $\begingroup$ @PieterB - Yes. You may have a point there! They say 'don't judge a book by its cover', but the fact is that books with nice covers get second looks. $\endgroup$ Oct 14, 2015 at 8:44
  • $\begingroup$ My assumption that the second puzzle might have been received somewhat better if you had taken only the last names, So Bush would have two peaks (making a better hint) and Ford would be mixed up with the car (causing a bit obfuscation). - I am not sure if "and they are getting tougher" was in any way intended as referring to the presidents instead of the puzzles (that would be very opinionated at least), but using any hintful and yet innocent words such as "lead" or "state" or "power" in the description would also have made it nicer $\endgroup$ Oct 14, 2015 at 14:05
  • 5
    $\begingroup$ To me this sounds like a meta question because it is about the reception the puzzles got on puzzling.SE, and about members' voting decision. $\endgroup$ Oct 14, 2015 at 15:53
  • 5
    $\begingroup$ @mikeTheLiar Certainly, questions about making puzzles are on-topic for puzzling.SE. But questions about making puzzling.SE posts belong on meta. This question is borderline, but to me the focus seems to be how the puzzling.SE community responded/will respond to a post. I can imagine some changes to this question that would make me think it is a better fit here than on meta. $\endgroup$ Oct 14, 2015 at 18:42
  • 1
    $\begingroup$ @Julian I think it's debatable, but as noted I'm just a tourist. I'm sure whatever the community decides will be fine :D $\endgroup$ Oct 14, 2015 at 18:44
  • 2
    $\begingroup$ We have a puzzle-creation tag for a reason, but I agree with @JulianRosen that this feels more like a question about the community's behavior than the puzzle itself. If it were something like "How do I make good ngram puzzles?", that'd be very on-topic. $\endgroup$
    – Bobson
    Oct 14, 2015 at 19:47
  • 2
    $\begingroup$ Speaking personally, I think there are arguments on both sides about where this should be. On balance I still think it belongs here. However I have now received plenty of excellent feedback and I don't mind if the question is migrated to Meta or even closed at this point. $\endgroup$ Oct 14, 2015 at 20:20
  • 2
    $\begingroup$ Personally I would like to see more puzzle creation questions, so if this does get closed as off-topic, my preference would be to see it revised and hopefully reopened here rather than moved to meta. $\endgroup$ Oct 14, 2015 at 22:19
  • $\begingroup$ It seems to me that the question is being called unsuitable because I related it to a specific case -- a series of puzzles based on ngrams. Of course I could make it more general and ask, "How can I make my puzzles more interesting?" but if we restricted ourselves to questions at that level of generality the possible topics would soon be exhausted. I really don't see what I could change to make it 'more suitable'. I had a specific question about one problem I'd encountered in puzzle-making, I asked for help, and I received some excellent pointers. What's to change? $\endgroup$ Oct 14, 2015 at 22:33
  • 1
    $\begingroup$ Naturally, this discussion would be better on Meta... $\endgroup$
    – user20
    Oct 14, 2015 at 22:46
  • 2
    $\begingroup$ @Emrakul - How do I vote to migrate this back to the main site? :) That option seems to only exist in the other direction... It's clearly now off topic and I should technically vote to close given: "This question does not appear to be about Puzzling Stack Exchange or the software that powers the Stack Exchange network" $\endgroup$
    – Alconja
    Oct 15, 2015 at 5:39
  • 3
    $\begingroup$ @Alconja VTC with a custom close reason, and when it's closed the migration will be rejected and it'll be sent back. $\endgroup$
    – user20
    Oct 15, 2015 at 6:42

3 Answers 3


To add to what @2012rcampion said, I think you hit the core issue when you said:

"Note that the questions don't have an identical format. The first question relied on a cryptic clue whereas the second require a deductive leap."

The first is clever and makes you smile when you see the answer, because you have that "aha!" moment when you get it, the second is basically a trivia question (what 4 things from one category correlate to these specific periods in history), and leaves no lasting impression, nor sense of accomplishment for the reader.

Ultimately, it seems to me that your fundamental issue is that you successfully came up with a clever puzzle, and incorrectly attributed that success to the ngram, when in actual fact, the cryptic part was the fun/rewarding bit. So when you made the second version, you took the ngram idea, but applied it to a fairly uninspiring trivia question and it unfortunately missed the mark.

Also as a minor aside, two other things that are related, though I don't think directly relevant in your case:

  1. People definitely get burned out by repeated/similar puzzles. If you look through the archives, you'll see plenty of people try to do puzzle series and the votes steadily drop with each subsequent one. That being said, I don't think this is the issue with your "series" as you've only posted two.
  2. Your first puzzle had an intriguing title which would have drawn many more eyes (look at the view counts, 1184 for the 1st & 75 for the second), and subsequently boosted the votes you received - though this boosting is only going to be an amplification, so also isn't the core issue you're facing.
  • 4
    $\begingroup$ About the views... The first hit HNQ, which often has an inflating effect on votes. Not only does it get more eyes, but many of those eyes aren't "regulars" here and don't vote the same way. Specifically, anyone with an association bonus can upvote, but only those with rep earned here can downvote. $\endgroup$
    – Set Big O
    Oct 14, 2015 at 3:24
  • 1
    $\begingroup$ @Geobits - I assumed as much about the HNQ status. But I think my point about having an interesting title goes hand in hand - i.e. if you land in the HNQ list, having an interesting title will make people much more likely to click through, which will amplify the amplification so to speak. $\endgroup$
    – Alconja
    Oct 14, 2015 at 3:33
  • $\begingroup$ Yes. I completely discounted the influence of the title -- a bad move. I think I was relying on previous success to carry the day. I suppose that is a common trap for anyone who enjoys a degree of success, in whatever field. $\endgroup$ Oct 14, 2015 at 8:48
  • 2
    $\begingroup$ @chaslyfromUK - yes, it's easy to underestimate the importance of flavour (of which the title is part). Case in point, check out the edit history of this post of mine. My original was getting down votes, so I edited it heavily. Same puzzle, different presentations, went from -2 to +29. $\endgroup$
    – Alconja
    Oct 14, 2015 at 13:14
  • 1
    $\begingroup$ While I think this explanation of voter behavior is entirely plausible, I think having a clever play on words should not be a cause for upvotes when it makes most of the puzzle at face value (the n-grams) become moot. $\endgroup$
    – xnor
    Oct 15, 2015 at 5:13
  • 1
    $\begingroup$ Related to your point that the second was essentially a trivia question, we are not all americans and relating things to the US Presidential cycles are seldom relevant outside US. $\endgroup$
    – Taemyr
    Oct 16, 2015 at 8:26

First a quick note: I found the second puzzle frustrating because I assumed that, like the first one, the solution involved some play on words. (In particular, I became convinced that the clue was "No 2" in the title, which I thought was hinting that the answers were common phrases missing "two," "to," and "too.")

As to why people downvoted it, I suspect that it was mostly for a similar reason: the lack of clues. The puzzle looks very much like (and perhaps is, a little) a "guess what I'm thinking" type of puzzle. My guess is that the reception would have been a little better if you had clarified that:

  1. We were supposed to guess the phrases based only upon the trends shown. Your hint "Where are the clues? They are all there. I trust the members of this site to find them." is far too ambiguous; in particular it seems to suggest that there are more clues hidden in the question (I tried looking at e.g. the second letter of every word, examining the post source to look for comments or other hidden clues, and even clues encoded in the two images). Your later edits pretty much fixed this one though.

  2. The four phrases share a common theme. Of course we know they have some relation to each other, but "four words or phrases that became widely-used in the latter twentieth century" is, in my opinion, a little broad. Something like "four members of a category" or "four items that share a property" might have been an appropriate hint.

Essentially I think the theme here is that at first glance the puzzle appears to be just arbitrarily difficult rather than clever. You need to let your readers know that the puzzle will be worth their time to try and solve.

  • $\begingroup$ Ironically the second was designed to be less of a 'guess what I'm thinking'. However, as you point out, I didn't make that clear from the outset. I actually took some time to find a suitable set of data. The idea was that the 'aha' moment would come from noticing that the peaks and drop-offs of the graphs were significant and formed a sequence. What I failed to do was sustain interest long enough for anyone even to bother to look for such clues. $\endgroup$ Oct 14, 2015 at 8:57
  • $\begingroup$ With reference to your point 2, it was initially quite difficult to think of data that would have a precise peak. It wasn't so much that they 'became popular'. It was that they became popular but then the popularity fell off again. Clearly in future I need to sell a puzzle and and not simply rely on the supposed goodwill created by a previous success. I aslo need to give some hint as to the method and not rely on the 'enigmatic' tag. $\endgroup$ Oct 14, 2015 at 8:57
  • 1
    $\begingroup$ I felt exactly the same way about the second puzzle as 2012rcampion describes in his post: I was looking for hidden words in the text, etc. I found nothing and decided to give up and to ignore the puzzle. (I only returned to it several hours later, after you had edited it.) All in all, the puzzle in its original formulation did not contain enough hints for starting to think about it. $\endgroup$
    – Gamow
    Oct 14, 2015 at 9:46
  • 1
    $\begingroup$ @chaslyfromUK I think you did not fail to create interest, or fail to hold my attention long enough (indeed, I spent almost an hour on it!) I guess the point I was trying to make is that a cryptic clue usually means a cryptic puzzle. You don't necessary need to hint at the method, but at a minimum let the reader know what type of puzzle it is. As I said before, your later edits fixed that aspect. $\endgroup$ Oct 14, 2015 at 17:33

I don't know why people downvoted one puzzle and upvoted the other. Voters are fickle. It could be bandwagoning. It could be, as your theorize, that reusing a format is a minus. I wouldn't worry that much about votes.

That said, I don't like either of the puzzles -- I was one of the downvoters on the first one.

The first one requires too much of a leap from the title, especially in Shakespeare -> William -> Bill. Someone did get it, but it strikes me as a case of many eyes. The n-gram data seems pretty useless. I don't know when the Kill Bill movies came out, nor would I have any way to think of it from the peaks without prompting. The puzzle emphasizes n-grams, but the real clue is the title, with the n-grams solely a final confirmation. It feels like a trap.

The second one, I was more OK with, choosing not to vote. Interpreting the data requires an insight and some history knowledge. Still, it's a bit arbitrary to guess US presidents (espsecially as you're Chasly from UK). It could have been some other set of historical leaders or events that happen in sequence, like a movie series. The mod-4 pattern is weak in the data. While the answer can be confirmed directly, it still feels like a solver would be unsure before having it confirmed, more of a "Maybe...?" than an "Aha!". I find that unsatisfying.

I like the n-grams idea, but think it needs a strong and striking pattern to make a good puzzle.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .