I Prefer Puzzles

Palindrome's team name this was "I Prefer Pi." Nametags designed by Joe Cabrera.

Palindrome’s team name this year was “I Prefer Pi.” Nametags designed by Joe Cabrera.

If you follow the MIT Mystery Hunt at all, then you already know that this year’s event, which ended Monday afternoon, was the longest of the modern era, taking an exorbitant 73 hours from the release of the first round of puzzles to the moment when a team finally claimed victory by finding a coin hidden on campus. Somewhere around the 48th hour of the Hunt, one of my teammates, a smart and good-natured fellow named Arun, called the Hunt organizers to ask for a hint about a puzzle.

Arun had basically dedicated his life to a particular puzzle called “In The Details,” and the description of this puzzle should give you an idea about the kind of thing we see at the Mystery Hunt. “In the Details” was a “fractal word seek.” After solving what you could on the first level of the word seek, you were expected (through some cryptographic and/or mathematical operation that I never did fully comprehend) to expand the word seek out to the next level of the fractal, and continue searching for words there. And then again, and then again. The people working on this puzzle ultimately expanded the word seek to the 85th or 86th level before locating all of the words they were expected to find. But even after all that, Arun and the others still didn’t have an answer to the puzzle itself. Solvers generally come away from a hunt puzzle with a word or phrase, which they will then need to use later on. Despite all the work that had been put into it, “In The Details” steadfastly refused to give up that answer.

So Arun called and asked for a hint. It was clear at that point that the Hunt was running long. Previous Hunts had sometimes ended by now, but this time, no end was in sight. The puzzles were extremely hard, and the meta puzzles even more so. A hint at this point seemed like a perfectly reasonable request. But Arun was dismissed with a firmness that I found rather surprising.

When my team, Palindrome, ran the 2008 Hunt, we too discovered we had overshot on the difficulty. Our response was to begin hinting as rapidly as possible, answering questions about the puzzles in great detail if need be. Why on earth couldn’t we get even a small nudge on this impossible word search? The only teams Palindrome refused to give hints to, years ago, were those teams in contention for the victory.

…Oh. And that is how I figured out that we were the lead team in the Hunt, or close to it.

Perhaps you suppose I might have guessed this, but really, I had no idea. The Hunt this year consisted of six distinct rounds — and at that point, we had only completed two of them. I thought we might be doing reasonably well compared to other teams, but winning? Or close to winning? That was hard to imagine.

On the other hand, we had a very large team this year. Palindrome is an open team: Anybody can join it. This year it felt like everybody did. Not too long ago we struggled to reach 50 teammates. This year we came mighty close to 100. Established team members brought in colleagues and friends. The Hunt organizers post a list of “unattached hunters” — people who want to play but who don’t have a team — and each year I pick up a few new people from this list. This year’s bunch of unattached hunters was a bonanza. I lured in a fellow named Eric. (We are required, it seems, to recruit a new Eric each year — we now have a great many of them.) This particular Eric, it turned out, was connected to a whole gang of Microsoft and Google engineers, all of them looking for a team. This turned out to be quite a bumper crop of excellent puzzling brains.

Also grabbed from the unattached hunters list was a total puzzle hunt newbie, a lawyer (I think) from Denver named Jonathan. Except he didn’t join Palindrome for himself. He came to the Mystery Hunt with his son, Brandon, who is 11 years old. His father told me that Brandon was a puzzle nut who couldn’t wait to one day participate in the Mystery Hunt. The trip to Cambridge was Brandon’s Christmas present.

I was a little nervous for both Brandon and his father — the Mystery Hunt can be overwhelming and difficult even for people who believe they enjoy puzzles, and if the Hunt did not match Brandon’s expectations, it could have been a disappointing trip home for them. I needn’t have worried. Jonathan told me that his son had been through the Hunt archive and knew what to expect. It would have been fairer to say that Brandon had memorized the Hunt archive. He could speak of themes and specific puzzles with an astonishing fluency, and he fit in with the team like a veteran. He was always busy with a puzzle, attended a bunch of events, and was still bopping along with the rest of us in the wee hours of the morning.

So Palindrome became a large team just in time to take on a Hunt seemingly designed for large teams. Still, it was hard to believe we were doing more than, say, keeping pace with the top third of Hunt teams. The vibe in headquarters was not “We are clearly killing this!” but rather “Help! We are drowning in puzzles! Very hard puzzles!”

Because the puzzles were indeed very hard. I think I understand how this happens. Unconsciously or overtly, puzzlemakers get it into their heads that they are creating puzzles that will be presented at the Massachusetts Institute of Technology!!!, one of the three or four greatest collections of smarts on the planet. It’s intimidating to consider the audience for these puzzles, and it takes great discipline to keep oneself from reaching for the difficulty knob, giving it a little twist, and then another little twist, until before you know it, you’ve cranked the difficulty all the way up to 11. All the while saying, “I hope these people don’t knock this off too easily…”

The metapuzzles at this year’s Hunt were particularly deadly. One of the great joys for advanced puzzlers is the joy of solving a metapuzzle even though you have not yet acquired all the answers that feed into it. When a metapuzzle has nine separate parts, and you discover the solution with only five of those parts, you feel like a puzzling giant. (Or so I imagine, as I have personally never accomplished this. But it’s fun even to be standing next to somebody who has just done it.) For the most part, the metapuzzles at this year’s Hunt did not allow for this sort of solving. Some of them were excellent, but others were assembled not from five or nine or fifteen separate answers but from dozens of individual puzzles — and those metapuzzles were structured in such a way as to force you solve nearly everything before you could really make headway on the solution. (Anyway, we certainly didn’t get anywhere with them. If you broke through on, say, the Feynman meta with only half the answers for it, I would sure like to know how you did it.)

Palindrome’s motto, as it says right on out nametags, is “We Play For Fun.” We’re not particularly filled with competitve spirit, and we’re not a good team for bloodthirsty Vince Lombardi types. I wouldn’t mind winning one day, but it’s more important that everybody enjoy themselves. Nonetheless, when I announced that we were one of the top teams in the Hunt, a subtle shift in tone could be detected. We didn’t go crazy with it — we didn’t tear up our nametags in order to get rid of our core motto — but you could tell that people were itching to grab the victory that we now understood to be within reach.

But there was a long way to go. We still only had two of the six rounds completed. There were so many puzzles floating around that it was impossible to keep track of them all, much less solve them. The hour grew late. We had already been through two graveyard shifts, and many of us had slept hardly at all through the weekend. Were we now going to stay up all night Sunday night — a good twelve hours after the Hunt would have ended in a more conventional year — to try to win this thing?


Teammates who had gone home came back in. Anya — a newcomer I had grabbed from the unattached hunter list but who had leapt into the weekend like she’d been on the team for a decade — brought homemade chocolate-chip cookies. (This was the Year of Cookies. I had rewarded graveyard shift solvers on Friday and Saturday with warm cookies from Insomnia Cookies. And another teammate, Ben, started us off with three giant containers of cookies, including something called Oreo Cheesecake Cookies. I don’t know if the Nobel Prize has a category for cookies, but if they did, Oreo Cheesecake Cookies would win.) Another teammate, Amy, brought the makings for waffles. A previously quiet young lady, Ange, transformed before our eyes: She whipped us into shape but good. She made extensive notes on what people were working on and what needed our attention, and she directed people like the world’s most efficient traffic cop. The Hunt organizers began allowing teams to buy puzzle answers at the rate of one per hour, and Ange kept time with the sternness of an SAT proctor. Rather brilliantly, she convinced the Hunt organizers that if we performed an act of amusing ritual humiliation, we would get an additional answer. To that end, we wrote and performed a pastiche of “Call Me, Maybe” (called “Solve Me, Maybe”).

And we solved puzzles. Lord, did we solve puzzles. I have been on Palindrome for close to two decades and I have never seen a burst of overnight productivity like what we accomplished this year, on Monday from midnight to five a.m. One by one, the puzzles fell, and then the metapuzzles. As the sun came up, we had only three puzzles left — the final puzzle for each of three remaining rounds.

We got stuck on all three of them.

The hours passed, and still there was no winner. The Hunt organizers announced that you no longer had to solve all six rounds — five would do. So now we only had to solve two of the three final puzzles. And we couldn’t do it.

– We had to transform a list of dates into an answer via a puzzle based in part on hieroglyphics. Some of the people working on this are professional puzzle creators; one of them had been the key figure in solving practically every metapuzzle all weekend long. We couldn’t do it. A thousand different approaches got us nowhere.

– We had to solve a puzzle involving a Rubik’s Cube — a puzzle so complicated I’m not even sure how to explain it to you. The people working on this included a physics professor from Harvard and a guy who, earlier that evening, had solved the Rubik’s Cube in about a minute, while simultaneously singing Tom Lehrer’s “The Elements.” We couldn’t do it. This team never really got stuck, per se, but their advancement through the puzzle was inch by painful inch.

– We had to assemble an Enigma machine from a cardboard tube and some pre-printed strips of paper. Some of the people we had working on this are employed by the National Security Agency. We couldn’t do it. Maybe we were overlooking something fundamental, after a hundred hours of brain-fry. I don’t know. But we couldn’t do it.

Soon the sun was in the sky again. People had planned for the Hunt to be over long before now — they needed to check out of their hotels, and get to the airport or to the train station. We had to clean our headquarters and convert the rooms back into classrooms. A large fraction of the team was sitting around waiting, hoping for a breakthrough that might get us back on track… a breakthrough that was looking less and less likely. Continuing the Hunt started to seem nuts — Captain Ahab territory. Finally, and painfully, I waved the white flag. If people wanted to continue solving, that was fine, but officially, we were done.

We spent an hour cleaning up. Somewhere along the way, the Rubik’s Cube solvers reached a solution. As I left to make my train, a small group of teammates decided to take on the one last puzzle we needed to solve before reaching the Hunt’s endgame. Amtrak was just pulling out of the station when I recieved the message: “We’re on the runaround!” Palindrome had reached the final stage of the Hunt after all. A few hours later, by the time I got home, I learned that we had completed the Hunt — a little over sixty minutes after the first-place team.

On the one hand, it’s hard not to ache at how close we came to victory. Our fingertips had brushed the brass ring, but we ran out of time before we could grasp it. On the other hand, at an event like this, a second-place showing is pretty darn impressive, especially in a difficult year like this one. I’m very proud of my teammates, who even when the Hunt was at its most frustrating, displayed nothing but encouragement and good humor (and waffles and chocolate-chip cookies). Plus, their puzzle-solving ability is nothing short of awesome, and I say that as someone who believes the word “awesome” is suffering from serious overuse.

Congrats to my friend Dan Katz, who was on the winning team. Dan has headed up many great Mystery Hunts, and I know that any hunt that involves his participation is a hunt worth looking forward to. As I write this, I am still very tired, despite having slept last night for thirteen hours. My feet and legs still hurt, though not as bad as they did yesterday. Full recovery is going to take a while. And yet somehow, I am already itching for next year.

Tweet about this on TwitterShare on FacebookShare on Google+Share on TumblrEmail this to someone
This entry was posted in Uncategorized. Bookmark the permalink. Post a comment or leave a trackback: Trackback URL.


  1. Posted January 22, 2013 at 6:14 pm | Permalink

    The meta structure of the Danny Ocean round was similarly broken, at least for us on Codex. Six different sub-metas, each of which was fed by 4 or 5 of the round puzzles’ answers–and each of the round puzzles fed into more than one sub-meta.

    On Codex, we had all but two of the puzzles in that round solved at one point. Unfortunately, all of the sub-metas included one or the other of those two puzzles. And with only 4-5 answers contributing to each of the metas, one missing piece of data was quite significant.


  2. Noah Snyder
    Posted January 22, 2013 at 6:42 pm | Permalink

    If you broke through on, say, the Feynman meta with only half the answers for it, I would sure like to know how you did it.

    We (Metaphysical Plant) got this one with 17/25 answers. That’s not half, but is still missing a lot. But the key thing was that we had the entire bottom half (though almost none of the top half), which was some very good luck (plus buying one answer). Also, this was Sunday afternoon, so we’d had a lot of time to look at it.


  3. Ryan
    Posted January 22, 2013 at 6:46 pm | Permalink

    So what exactly was the timeline for the end? We got an email saying that Atlas Shrugged won at 11:56. But apparently the coin wasn’t found until 3:21. If you guys were only an hour behind, and the runaround involved a whole bunch of puzzles, how could they call it so early? And how did Palindrome finish an hour behind, when the wrap-up was already occurring at the point this was stated?

    @codeman38: Yeah, we didn’t even bother to try many of the casino puzzles–and thus the meta–until late Saturday. The ability to buy a few answers helped us out a lot. Of course, then we got stuck on the meta for 24 hours until they gave us a hint…

    Up Late


  4. Ryan
    Posted January 22, 2013 at 6:52 pm | Permalink

    @Noah: It looks like we at Up Late got Feynman with maybe 18/25 (rough count of what I can see at the moment). I think that we too had mostly the lower grouping. No free answers here…we used them to buy half of Get Smart. :)


  5. Eric Berlin
    Posted January 22, 2013 at 7:04 pm | Permalink

    Re: Danny Ocean: It’s hard to imagine how that would have gone had Foggy Brume not been on our team. He made a spectacular without-a-net leap to the right idea when we had holes in all but one of the smaller rounds. It was one of the more impressive feats I witnessed this weekend, although one cannot discount the Elements/Rubik’s Cube thing.

    As for the timeline of Palindrome’s finish: I’m not sure. I was most of the way home when it happened, and I never even glanced at the wrap-up announcement as there was no hope that I would be there. Perhaps a few representatives Manic Sages stayed behind to let Palindrome finish. (If that is the case, then that was a lovely gesture.) And perhaps Palindrome finished the endgame slightly faster than Atlas, allowing us to catch up a little on the clock. That’s all just conjecture, though. Perhaps someone who was there can comment.


  6. David Glasser
    Posted January 22, 2013 at 8:01 pm | Permalink

    Nice to run into you on the train, Eric!

    Even if my screams of understanding when you mentioned the “RED HERRING” part of Time Conundrum (which we somehow managed to solve without that big piece of information) may have woken up several sleepy folks nearby :)


  7. Posted January 23, 2013 at 1:19 am | Permalink

    eric, fantastic writeup, as always. you deserve massive, massive credit for your leadership this weekend. i am convinced that the main reason palindrome performed as well as we did was superior morale. that was pretty much the only thing carrying us through some of the bleaker times (which included, let’s say, all 24 hours of sunday, during which we solved zero metas and i personally was a useless sack of bleagh).

    as the “harvard physics professor” (actually a significant exaggeration, but it’s a common misunderstanding of my job title) mentioned above, i can explain the rubik supermeta… even though my teammates really had the huge breakthrough only after i finally went to bed at 9 am monday. (from glancing around the google doc, i can infer that sages’ yes/no hints were critically useful here.) the puzzle was called cryptocube, and indeed, it was a cryptogram in which each letter mapped to a particular move on a rubik’s cube. of course, they didn’t actually come out and say this; as with seemingly every puzzle and every meta, there were no instructions and no flavortext. the puzzle merely showed 26 different* words, each with a picture of a particular state of the cube. we inferred that each of those states was the final state after applying, in order, the sequence of moves corresponding to each letter of the code word, starting from a solved cube. so the task was, obviously, to figure out which move corresponded to which letter. if we ever got to that stage, it was fairly clear what we needed to do to our meta answers in order to get, if not the final extraction (wishful thinking, right?), at least a clear next instruction.

    so that is already a brutally difficult puzzle. this being hunt 2013, they turned it up to 11 by including in the list two instances of the code word LAUGH (hence the * above)… corresponding to two slightly different cube states. whaaat? nevertheless, APH had emailed out the hint “don’t let LAUGH dissuade you from an otherwise promising idea”, so … we didn’t. i actually theorized that there was an invisible 27th crypto token (e.g. a space) to complete the set, since there are 27 rubik moves (if you count a center-layer twist as a single move rather than treating it as an opposite twist of the two faces that sandwich the layer) and only 26 letters. that would have been outrageously devious, but at this point, 60+ hours into the hunt, i was fairly convinced that that was what was going on—the way the numbers worked out seemed really promising.

    the actual explanation turned out to be, if anything, even more ridiculous: you had to drop an unspecified letter from each code word before applying the moves. so the two LAUGHs were different because a different letter had been dropped. wow. my teammates actually figured this out (!!) and, ordering the clues by which letter had been dropped, extracted an alternate set of instructions indicating the order in which to apply the moves corresponding to the meta answers. palindrome was the only team in the hunt to solve this meta. huge props to everyone who brought it home after i left.

    i really couldn’t be happier with second place. next year, though? we could win this. right?


  8. Dan
    Posted January 23, 2013 at 3:51 am | Permalink

    Yeah, we totally underestimated the difficulty of our puzzles. Massively so. Really, I think what happened is that we underestimated our testsolvers. Thanks to all who plowed onwards despite the difficulty.

    The reason we sent out the “hunt is over” e-mail when [Atlas Shrugged] solved their fifth meta was that so many people were so tired that we felt we had to get the word out. I don’t think it was feasible for Palindrome to catch up on the runaround anyway (unless they’d made a mistake on an obstacle, which they didn’t) and I think the rest of Hunt appreciated the extra few hours of “okay, guys, go home.” (That said, we continued to receive call-ins even into wrap-up!)

    When Palindrome made the runaround, we were going to lose room renovations and we’d been planning to strike the obstacles as [Atlas Shrugged] got through them. We’d only struck one or two obstacles, so we halted that plan and then ran as much of the runaround as we could for Palindrome. I’m glad: we wanted more people to see it!

    Thanks all — I hope you’ve gotten some rest!


  9. Eli
    Posted January 23, 2013 at 8:44 am | Permalink

    For me, the difficulty level of the extractions (I feel like I spent most of my time staring at full tables of data trying to make that next leap) had the effect of deterring me from even attempting things later on. For example, the Dropquote. I looked at it and thought “all right if I spend two hours making some sort of sense out of this, what are the chances I’ll be able to actually extract something?” I’m sure I’ll look back at the solutions and be impressed, but for me it just wasn’t compelling.

    That being said, it was Mystery Hunt for god’s sake. Just being there makes my year. And it was Palindrome for god’s sake. Impossible not to have a good time just hanging out with those people.


  10. Posted January 23, 2013 at 10:53 am | Permalink

    @Eli: YES. This was another issue that Codex kept running into: extraction methods that were extremely under-clued.

    It seemed like there was a ridiculously high number of puzzles where we’d figure out the initial “aha” fairly easily, get all (or nearly all) of the relevant pieces of information needed for extraction…and then we’d just get stuck, fruitlessly wasting hours on figuring out the second “aha” of how to actually extract anything from them. (The two that particularly come to mind for this one were “Mashup” and “Plead the Fifth”.)

    And then there was the error by Sages on one puzzle that further exacerbated this. We had managed to figure out an extraction method for that puzzle and called in the answer “PLANAR”…and were told it was wrong. So we ended up going on a several-hour-long chase to figure out what we could manage to extract instead of “PLANAR”, to no avail (I think we were even considering calling in its homonym “PLANER” at one point!)…until someone had the idea to just call in “PLANAR” again, at which point it was ruled correct.


  11. Mark Halpin
    Posted January 23, 2013 at 4:12 pm | Permalink

    We also solved the Ocean’s 11 meta by taking a large flyer on how letters from sub-answers were “stolen” from “casino” answers. Looking at the actual solution earlier today, it’s impossible for me to imagine anyone solving it as intended (and as someone whose own puzzles are almost always a little harder than I think they are, I have an extremely fertile imagination).


  12. Posted January 23, 2013 at 4:42 pm | Permalink

    I think we got Feynman with 15 or 16 of the 25 answers. We knew the process, had backsolved YANKS and gotten close to backsolving POLIO but prefered “BORON” as the answer there. But then it was just cranking the few close places we had. With 7 or 8 of the letters in the message and an idea for the enumeration, Wheel of Fortune killed the round dead. It was my happiest moment at the Hunt, as it was our first meta to solve and it ended a lot of puzzles by being a 60% completion meta.

    Nothing else was that low for us this year, and metas should always be solvable with 80% data or so just because # broken puzzles will rarely be 0 in a Hunt of any size, certainly this size.


  13. Arrchivist
    Posted January 23, 2013 at 7:41 pm | Permalink

    Not to mention that we were the only team that solved the Rubik’s Cube supermeta-puzzle, at least according to the people at wrap-up. Also, we were the only other team than the winning team to get to the runaround by solving 5 supermetas. The next top number was 3.


  14. spate
    Posted January 24, 2013 at 12:06 pm | Permalink

    Hey All,

    Just wanted to say how proud I am to be able to join Palindrome. The welcome you all showed to someone who was not all that helpful was amazing! I certainly plan to join again next year, will be practicing so I can pull my weight!

    Thank you Eric!



  15. Joe
    Posted January 24, 2013 at 9:13 pm | Permalink

    As the guy who can sing The Elements while solving the cube (a talent I hope to do on a resurrection of the Gong Show someday), I’d say that that puzzle kind of seemed that it was made for people like me. In fact, I’ve actually given a few talks about the group theory involved in solving the cube.

    I went to grad school with Eric Wofsey, who is listed as a designer, and can verify that in spite of being the youngest person in our program, he was the smartest, by a long shot (this includes most of the professors). I’m really ecstatic to have solved it and to have been the only team to do so. And to be honest, I can’t think of any serious problem with it other than the requirement of having all six meta answers to solve it.

    I’d also like to publicly thank my brother, Jason, for being willing to help us with this via Skype at 7 AM on a federal holiday.

    Regarding the fractal word search, I believe it was noticed that there were 2×2 squares in the grid that repeated, indicating that they were starting on “Level 2”.

    As for the other puzzles, I seemed to have particularly bad luck working on ones that had a simple structure, required a ton of work then came to a dead end (Circuit Board, Infinite Cryptogram, etc).


  16. Amos
    Posted January 25, 2013 at 3:32 am | Permalink

    Joe, thank you for this comment: “And to be honest, I can’t think of any serious problem with it other than the requirement of having all six meta answers to solve it.” Also, thank you to your team for solving it, I would have been rather sad if no one had.

    Eric and I designed the majority of the Rubik round together, and I ended up writing Cryptocube itself. We agreed during development that it was bad to require all 6 regular metas to solve it, and at one point the round was cut because of that (and because the actual construction of the overlapping metas was really hard).

    Eventually we needed a 6th round still (in retrospect the wrong move), and so we provided the essentially check-sum that should’ve made it theoretically possible to solve with only 5 out of 6, though we never actually got a chance to test it as such. Thinking back, the check-sum should not have used meta answers alphabetically, and should have provided the actual ordering instead.


Post a Comment

Your email is never published nor shared. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

  • Archives

  • Subscribe

    By signing up, you agree to our Terms of Service and Privacy Policy.

  • Twitter