I recently read Tony Bourdain's Medium Raw, which was a fascinating look into the world of people who are really, really interested in food. I like food. These people think about food more than I do. So much so that I can barely understand their explanations.
At my first meal at Momofuku Ssäm, one particular dish slapped me upside the head [...] It was a riff on a classic French salad of frisée aux lardons: a respectful version of the bistro staple -- smallish, garnished with puffy fried chicharrones of pork skin instead of the usual bacon, and topped with a wonderfully runny, perfectly poached quail egg. Good enough [...] But the salad sat on top of a wildly incongruous stew of spicy, Korean-style tripe -- and it was, well, it was... genius. Here, on the one hand, was everything I usually hate about modern cooking -- and in one bowl, no less. It was "fusion" -- in the sense that it combined a perfectly good European classic with Asian ingredients and preparation. It was post-modern and contained my least favorite ingredient these days: irony. [...] But this was truly audacious. It was fucking delicious. And it had tripe in it.
(--from Medium Raw, Anthony Bourdain, chapter 17)
Mind you, the whole book isn't like that. Bourdain talks about everything from hamburgers, to fatherhood, to foie gras, to the Food Network, to the stupid things he wrote in his first book. But that paragraph in particular grabbed me because I have no idea what he's talking about. I can look up the recipe (frisée lettuce with hot pork, vinaigrette); maybe I've even eaten it somewhere. I've eaten spicy Korean stews. But why is this ironic? Or audacious? What is it reacting against? What are the things it is reacting against reacting against? If I'd been sitting next to Bourdain, eating off his plate, I still wouldn't have a clue.
I recalled this paragraph on Sunday afternoon, sitting in an MIT auditorium, listening to the designers of the 2011 Mystery Hunt talk about their puzzle structures. I knew exactly what they were talking about. I'd just lived through it (or half of it, anyway, since I got two good nights' sleep during the Hunt.) Everybody in the room was smiling and nodding along to the speaker's presentation, and laughing at the jokes on the slides. This was our field. This was our side of the wall. Tony Bourdain would have been completely befuddled, see?
(Mind you, if I'd tried his salad, I'm certain I would have enjoyed the hell out of it. Puzzles have more of an entry barrier. But put that aside.)
I want to talk about how the Mystery Hunt has evolved in the five years I've experienced it. But that wouldn't be enough perspective. Many of my teammates have been doing it for ten years; some longer. Several of them have designed Hunts. Somebody needs to write the Hunt history thesis and it shouldn't be me. But I can start pointing at the questions.
What was new in 2011? What does puzzledom look like when it's playing above itself, reacting to things that the non-puzzlers have never heard of? I'll put down two lines, and then fill in the explanations for those of you who don't know from lardons.
Backsolving is solving.
Metas are a tool.
The earliest Hunts, we are told, were unstructured lists of puzzle questions. Then some genius added the idea of the "meta", or metapuzzle -- a puzzle built using the answers to other puzzles. (I first encounted this concept in The Fool's Errand, in 1988 or so.)
A simple example (not from any Hunt or game in particular): imagine you've solved a group of ten puzzles. The answer to each is a ten-letter word or phrase. In fact, each answer is a ten-letter name, and it's the name of a famous scientist or inventor. ("MARIE CURIE", for example.) You write down the ten names, in order of their famous discoveries (radium, 1898). That gives you a neat ten-by-ten-letter square. Then you read down the diagonal of the square. It spells out a new ten-letter word, which is the answer to the metapuzzle.
(Why the diagonal? It's not an arbitrary gimmick, although it is something of a genre convention. You need to pull an answer out of the letters of ten names. The important insight is that the order of discoveries is important. Given an ordering, you can pull the first letter from the first name, the second letter from the second name, and so on. The diagonal is just a way to visualize this rule.)
(Why not simply use the first letter of every name? Some metapuzzles do work that way. It's a question of puzzle difficulty. No insight is needed to look at the first letters -- that's such a common convention that we do it automatically. With that setup, you don't have to figure out the ordering of the names. You do have to unscramble the letters, but a ten-letter anagram is trivial with the right software. So that would be an easier final stage, which the designer might use if the earlier parts of the puzzle were particularly hard.)
(By the way, this example is kind of weak -- Marie Curie discovered more than one thing, you know! And radium could be said to have been discovered in 1898, when it was identified, or 1910, when it was isolated in pure form. A serious puzzle designer would eliminate these ambiguities. Fortunately, I'm just making stuff up for a blog post.)
But this metapuzzle system leads to an interesting side effect. You can solve a meta without solving all the puzzles that feed into it. If you've solved nine of the round's puzzles, figured out the ordering, and gotten "INS-GHTFUL", you don't need to solve "MARIE CURIE" to guess that last letter. You punch in the meta's answer and move on to the next round.
That leads directly to the question of backsolving. Say you're in this position, with nine puzzles solved. You can easily solve the meta; but you also have a lot of extra information about the earlier puzzle, the one you're missing. Because of the meta structure, you know that it's a ten-letter name, a famous scientist. The fourth letter of the name is "I"; and the scientist worked between, say, 1880 and 1915 (or whatever the years of the third and fifth letters were). With that information (and Wikipedia) you could probably guess "MARIE CURIE". That's backsolving the puzzle -- working from meta-information you know about the answer.
So do you punch that backsolved answer in? In my first hunts, my team preferred not to. It seemed like a form of cheating, and it didn't really get us anything -- not when we already had the meta solved. (The winners aren't the team that solves the most puzzles; they're the first team to solve the last puzzle.)
But this year, the organizers made a couple of subtle changes which flipped this on its head. First, they used a point system in which solving any puzzle got you closer to unlocking new puzzles. (Thus, going back to fill in old gaps was valuable.) And second, they added a simple checkbox to the answer page: "Did you backsolve this puzzle?" Just by recognizing that option, they made it feel more legitimate.
As a result, everybody did a heck of a lot more backsolving this year. And my impression is that this generated more fun for everybody.
After all, any given Hunt puzzle involves looking for patterns, and working both backwards and forwards between the clues and the answers. (If this makes no sense to you, think about crosswords. Of course you work back and forth between the clues and the grid. Looking at the crossing letters in the grid isn't cheating, it's the whole point.)
If metas are part of the solution process, then that back-and-forth information flow becomes multilayered. Any puzzle might require both clues and context to solve. That can only lead to more interesting puzzles.
(Plus, of course, backsolving is solving, and solving is fun. One teammate remarked that the best two moments of the weekend were the Hunt's launch, when the first brand-new puzzles appeared -- and 3 AM Sunday morning, after the successful cracking of a meta pulled the group into an intense burst of fruitful work on its related puzzles.)
Back to this year's metapuzzles. Metas are now a standard Hunt element. Standard enough, in fact, that for several years everyone took them for granted. That's why "metapuzzle" got abbreviated to "meta", right? A round consisted of a bunch of puzzles and a meta. Solve all the metas, you get into the endgame. That's the way my first Hunt worked.
There were always variations in this structure, of course. But the 2011 hunt got a little more crazy than usual. It was divided into five rounds -- five "worlds", as it had a videogame theme. Each round was roughly twenty puzzles, divided into (say) three groups. Each group had a meta. The solutions to the three metas then had to be assembled into a meta-metapuzzle for the round. When you had the five meta-metas, you got to the endgame. (Which was not technically a meta-meta-meta, because you weren't assembling the five meta-metas into a new answer -- you just had to collect them.)
Furthermore, each of the five worlds had a different meta structure. (Spoilers coming for anyone who wants to try the Hunt puzzles...) The first world was an unadorned meta-meta, involving the answers to the three metas. In the second world, each meta answer describes a transformation that has to be applied to the previous meta puzzle name. In the third round, each puzzle has three answers, one for each of the three metas... and so on.
The creators were justifiably smug about their experimentation. In a sense, they wrote five mini-Hunts, each with a creatively different meta structure.
In another sense, I think, they put the knife in "the meta" as a concept. (Although it may be a while before it expires.) The meta-metas are the first hint. Why not go for a meta-meta-meta? Well, you could, but it wouldn't be three times as clever -- it would just be another puzzle relation. These new structures? They're interesting puzzles, which involve the answers to other puzzles. But all the puzzles in a Hunt should be interesting! And they are.
Metas aren't qualitatively different puzzles. They're a tool for hooking puzzles together.
For some reason, the Hunt spent several years going around in this loop where all the metas kind of looked the same. I mean, they were distinct puzzles -- but they all had the same shape, where you looked at an incoming set of puzzle answers and applied brain-sweat. This, as the non-meta puzzles went through cycle after cycle of creative improvisation.
So what will the Hunt look like in 2015?
I'm waving my hands, of course, and I don't particularly expect to be the one writing the 2015 Hunt. (My team consistently does well, but not that well.) But let's say that I'm right.
Designers will let go of the notion that solving has a direction. It won't be puzzles feeding into metas; it'll be puzzles connected to other puzzles. And those puzzles connected to yet others, or maybe back to the first bunch.
I don't mean that all structure will dissolve into an N-simplex of every puzzle using answers from every other puzzle. That wouldn't be constructible. (Although someone's gotta try it.) But the point of puzzle design is insight, and insight about where the answers are coming from -- or going to -- is a valid dimension to play with.
Certainly there will be puzzles which you pick up and solve directly, with no other context. My points are (a) knowing that from the start is less fun than figuring it out; (b) if you are stuck, shouldn't the context -- the related puzzles -- be an avenue of attack? Start wherever you can make an entrée, and work around to the puzzles you're stuck on. The more directions you can approach your stuck spots from, the more fun you'll have.
Maybe a "nexus" puzzle -- one which involves the answers to many other puzzles -- will actually be solvable on its own. (Or partially so.) You'd be expected to then "backsolve" into the other puzzles and make progress on them from there. Or maybe you'd find two puzzles that were unsolvable on their own; you'd have to work them in parallel, through a common relative.
Maybe all puzzles will just get a little bit harder, because back-and-forth is the expected mode of solving. That might be frustrating at times, but then it would be more satisfying at other times. It's hard to deny that Hunts have been getting more complex, though the creators have tried to keep the difficulty balanced. This sort of interweaving is the way I see that evolving.
And, in closing, I'll link to the 2011 Hunt's opening act and closing credits. If this all sounds like tripe, you can still watch those.
Mmm. Spicy Korean tripe.
I've been trying to come up with a useful comment but I can't add to this awesome analysis in any way, other than thanking you for sharing it.
Oh, my word, there's so much to discuss here. So much, in fact, that it's probably easier to meet over coffee (or at least something that coffee is metaphorical for) to discuss it.
Because in a sense, what you say is incorrect. There are reasons there are metas; there are reasons puzzles don't interlock, or interact, except perhaps in some carefully defined ways. In 2002, when we were writing the Matrix Hunt, we had "training puzzles": in order to finish a puzzle "inside" the matrix, you needed to call on some training (a la "Whoa. I know kung fu!") that you got from another puzzle. One question that arose was this: should that training puzzle make it possible to solve an otherwise impossible puzzle (so that, say, your "kung fu" training told you that clues with prime numbers needed to be ignored, or reversed, or whatever)? Or should a "inside the matrix" puzzle be solvable on its own, with training applied only to its answer to get you a new answer that you actually needed for the metapuzzle? In the end, our team leader insisted that teams shouldn't pick up a puzzle and find out that it's unsolvable; that's really just not fair to them. Instead, you should get to an actual answer (e.g., DOUBLET) and then use your training ("take out the blue") to get a new answer (e.g., remove BLUE from DOUBLET to get DOT). Puzzles, the theory went, shouldn't depend on other puzzles.
Similarly, there are reasons there are rounds of puzzles. In both Zyzzlvaria and this Hunt, the first section (the "inner board"; the Mario worlds) were a sort of self-contained small hunt, so that teams of more casual solvers could reach a point that felt like an accomplishment by solving that small hunt. My wife, earlier today, remarked that some year the Hunt structure would involve giving teams all 20 puzzles on Friday at noon, still designed to take 36-48 hours; I pointed out that for a smaller, more casual team, that would be nightmarishly intractable.
So as I said, in a sense, what you say is incorrect. And yet, as that implies, in a sense you're correct, because the entire point of (at least some) innovation is to take the things that everyone knows can't be done, and then do them anyway. The Hunt of Evil had seven puzzles that were more or less obviously unsolvable, like an underdetermined sudoku and a really boring "spot the differences"; those became solvable when you got the other half of the instructions (the fact that the sudoku never had two prime numbers adjacent; the nearly-identical spot-the-differences page that you had to compare to the original spot-the-differences page to spot the actually relevant differences). The Reverse Zone of the 2009 Hunt had "Doctor" puzzles that were utterly undoable until you solved the relevant "Companion" puzzle, which gave you the information to solve the Doctor puzzle, and you combined the two answers to get the actual answer you needed. I think Acme's leader might have found those unthinkable.
So the things that I know are obviously necessary are exactly the things that the innovators will be ignoring by 2015. All the same, there are certain guidelines that they'll presumably have to work within--there's breaking expectations, and then there's breaking conventions ("read the first letters" and "read the diagonal" are, as you note, conventions; a puzzle where you get the answer by reading the second letters, unless the puzzle is themed around famous seconds like John Adams and Buzz Aldrin and Exodus, would only annoy people). Separating the unbreakable conventions from the wholly breakable traditions is...difficult. And interesting. And we'll see what happens.
I've got entirely other things to say about backsolving (some of which I said in the comments on a post about last year's hunt, which I linked from my writeup of this year's hunt, which is my Dreamwidth journal, username "tahnan"), but this comment is already reaching epic length as comments go, so maybe I'd better stop.
I had a good time here but will return to google now.
I have now survived another Hunt. I'm not sure what I think about this stuff any more, but I want to pass along this:
As history will record, the 2013 Mystery Hunt was way too fricken' hard. (It was the longest Hunt on record -- something like 72 hours before a team reached the end, and that was *with* the copious hints and free solutions that the designers started handing out halfway through.)
The designers said they wanted to create a Hunt with no backsolving. Chris Cieslik, on our team, made this comment:
"a parting bit of analysis... the Sages were successful in eliminating backsolving... but in doing so the hunt became uncompletable -- and they were forced to let us unlock puzzles for free... the ones we would have likely backsolved. so backsolving is a natural part of puzzlesolving that culls the weak puzzles from the pack."