A giant vat of molasses. Image by Setec Astronomy.

What I did at Mystery Hunt 2019

Here’s my “What I did at Mystery Hunt 2019” blog post.

I’m assuming that if you’re reading this, you know what puzzle hunts are, you know what the MIT Mystery Hunt is, and you know how the structure of the 2019 Hunt worked; otherwise, go read some other blogs and come back here. A lot of the data below is a mix of my memory and our puzzler-tracking system; if some puzzlers weren’t engaging with the system correctly, my numbers below might be off.

About Me

I was on Team Left Out, the team that ended up winning. We had 59 team members; 39 playing on-site at Boston, 15 playing from California. Our team has been playing since the 2005 Hunt. This is the only team I’ve ever played with. I’ve played with them every year starting with the 2006 Hunt. I’ve played from California 13 years in a row, but this year was the first year I played on-site at Boston.

General Stats

Throughout the 54-hour event, I logged time working at 56 puzzles. 18 of them were quick “oh, I don’t think I can help with this more than what’s already been done by the team.” The other 38 puzzles, I’ve written a quick summary of what I worked on below. There were 181 puzzles (counting events and metas) total, so this means I made some sort of substantial contribution to about 1 out of every 5 puzzles. Based on my reading of other people’s blogs, that might be a bit on the high side compared to the average solver.

Some notes about my summaries:

  • Potential spoilers have been encoded with ROT13.
  • For crediting purposes, it’s sometimes the case that multiple people got the insight at nearly the same time, so when I claim I got the insight below, it’s not always the case that I was the only person on the team to do so (although it often is).
  • “Grinding” is the term I use for the crank-turning bits of the puzzle. This includes looking up data, solving sub-puzzles, typing text into spreadsheets, etc.
  • “CSGP” is short for “Constraint Satisfaction Grid Puzzle” (or rather, CSGP-heavy, since in Hunt usually it’s more than just the CGSP part). These are the type of puzzles often called “WPC-style” in the overall hunt community — often they are culture-neutral, lots of visual logic. I don’t like calling them “WPC-style” because I feel that the WPC would be better of branching into other puzzle types, but that’s a tough uphill battle. Anyway, these often end up in my lap on my team just because I’m rather fast at them — although believe it or not, I’m generally not the fastest on the team, as that honor goes to Will Blatt.
  • I know a bunch of readers here might be looking for puzzle recommendations. The problem is, I enjoyed all the puzzles here, and I hate picking favorites. So, you’re on your own for that.
  • The puzzles are organized by round, not by the order I worked on them.

Towns (Events)

Flag Day: I attended this event.

Christmas

Ornaments: Did some grinding.

Halloween

Costume Change: Did some grinding.

Lantern Festival: CSGP. Did the grinding with 2 other solvers. The insight for extraction requires two parts; the first part was done by Mike Springer before I could get to it; I got the insight for the other part.

Send Yourself Swanlumps: Got the first message (while walking back from picking up the artifact)! Did the grinding with 2 other solvers. Totally overthought the first message, but eventually got the insight as to how to interpret it. Got the final insight and the extraction.

(It’s rather serendipitous that I ended up being the main person on the team to work on this puzzle, as I was also instrumental in creating a very similar puzzle (in presentation, at least) for the Doctor When Game. Non-objectively, I think my version of the puzzle is cleaner and has more interesting stuff, but YMMV. If you’d like to try it, here are PDF links for the cover and contents.)

New Year’s Day

Chowing Down: Did the grinding with 2 other solvers, although we sometimes asked the room for help. Got all the insights.

Far Out : Did not get the initial break-in. Solved it on paper with 3 other solvers. The final extraction wasn’t much of an insight.

Twelve! Eleven! : 11 solvers jumped on this grind-fest. I’m pretty sure I got the insights along with someone (everyone?) else.

Valentine’s Day

Comma and a Freaking Dot: Got the initial insight. 4 teammates jumped on to do the grinding, but I left to do some other puzzle. Came back and got the extraction. Other teammates eventually did the creative portion — and then nobody bothered to follow-up on the reward for doing the creative portion. So I did the audio extraction and last part by myself (although I did ask the room for some ID once I had the final image). I apologize to the question writers for not being a fan of the band being referenced and therefore totally missing all the references.

Fake IDs: Jumped on a team of 3 (down from 7) after most of the data harvesting had been done, but shortly after someone else had the first insight. Did a bit of grinding. Someone else did the extraction.

Invisible Walls : CSGP. Worked with 2 other teammates. Did half of the grinding, got the insight, did the visualization work to get the final extraction.

Arbor Day

Complimentary Copies: Did the grinding with 4 other solvers. Got the insight and the extraction.

Flocks: Got the starting insight. Did the grinding with 2 other solvers. Did not get the extraction.

Loaded: Did the grinding with 3 other solvers. Got the overall insight and some of the small ones. Got the extraction.

Touring the Nation: (Oops, forgot to add this one on the first version of this post.) It’s rare to get to fully-solve a cryptic in Mystery Hunt — usually you do just enough to get the extraction and you leave the rest of the fun behind to go work on other puzzles. Here I worked with four other teammates, we got the first couple of messages, I went, “hey, I know these people!”, and then we did the other puzzle… and then we were stuck for what felt like forever, but was probably more like 10 minutes, until I got the last insight.

Thanksgiving

Herbert West, Animator: Assisted in some clean-up grinding.

Jukebox Hero: Got the early-to-mid insights, did the grinding with about 10 other teammates. The final extraction step was a bit awkward for us. I don’t think I was the one who got it, but it’s hard to tell.

You’re Gonna Need a Bigger Gravy Boat: Worked on this with 3 others. The bottleneck was pretty much cutting out the pieces, after that the puzzle pretty much flew. I think I got all the insights on this one.

President’s Day

Helvetica is Only an Okay Font: I jumped in, knew immediately what to do, and knew immediately that I didn’t want to do it. I did about one-third of all the letters, but I don’t actually think that was much help. I feel somewhat bad at abandoning Jesse Morris to grind on this puzzle.

Insider Trading: I jumped in when two of the groups were solved but before the meaning of “trading” was understood. Wow, this was some really really slow grinding; it’s like a separate “aha” was needed for every group. I left when there were seven unidentified groups, then came back when there were four. Stuck around until there were two, and then I found the extraction mechanism by noticing “EROFEB” in one of our data streams. It still almost needed the last two letters before we could solve it, though. I feel very embarrased at not getting the last group since it was identical to a group that I failed to get in Puzzle Boat 5 just a week ago.

No Shirt: I started on this by copy-pasting data into the shared spreadsheet, and typing in enumerations and blanks. By then, three teammates had joined me, but they were all in California. Someone (not me) got the insight, but my search-kung-fu meant that I did about half of the grinding. We got an extracted message with two letters missing… and then, stuck at the message F?L?ASIANISPELK. With not much else to do, we spent another 15 minutes getting the last two letters, and that’s when I left the puzzle. It was solved 15 minutes later, when someone discovered that the spreadsheet had wrong indexing data — totally my fault! — and that the first letter was wrong. Amazing how one wrong letter can lead to 30 minutes of wasted time.

Holi

Chris Chros: Joined up with 2 teammates who already had the insight and a lot of the grinding. Did 1.5 of the four grids. Extraction was trivial.

Connect Four: CS(G)P. Joined up with 4 teammates who were making slow progress on the first step. Got the insight for the first message (not that hard). Totally missed the relevant anagram that was supposed to be a hint that you were on the right track (not that important). Did pretty much all the grinding for the second step. Did not get the final insight/extraction (fortunately, someone else did eventually).

Bloomsday

Bloom Filter: Guessed some flower names correctly. Puzzle ended up getting backsolved.

Concrete: Did the whole puzzle solo, with a tiny bit of quick ID help from my wife.

Schematics: CSGP. Did the grinding with 2 other solvers. Got the first extraction. Did the second layer of grinding. Did not get the final extraction.

Patriot’s Day

Fake Estate : I started on this solo. Two teammates from California joined up. I got the first message and realized an artifact was needed. I got the artifact and started typing in data that would be harder for the folks in California to do. A local teammate joined. And then… we opened up Movie Marathon and I was yanked over there because of expertise. This puzzle was solved while I was still on Movie Marathon.

Movie Marathon : This was probably the puzzle I spent the most time on in the hunt, a whopping 3.5 hours. That happens when the puzzle is just a 45-minute video. That you have to harvest a bunch of data from, and you’re not sure if there’s data that you might be missing. I got most of the insights, but missed a rather critical one — so as a result, we had a message we couldn’t understand how to interpret, and chased down lots of dead ends. After spending 2.5 hours on it, I wrote down our best guesses and some notes, marked the puzzle as “needsaha”, and moved on. Seven hours later, Nina Hinrichs noticed something all the movies seemed to have in common. I went “Oh carp”, came back, did another hour of grinding, and solved the puzzle.

Mathematical! : This was the penultimate regular puzzle to be solved. It was reasonably obvious that this puzzle with 30 different video clips and a binary diagram had something to do with Gray code or the word “Gray”, something to do with TV shows, and something to do with commemorating the past, and something to do with adventures.

I had an insight that covered all of this. Namely, that in the second season of Star Trek: The Next Generation (a sci-fi adventure TV show) there is a clip show (a show that is mostly recaps of previous episodes) called “Shades of Gray”! In that episode, they show 37 clips from previous episodes. The puzzle shows 30 clips of different TV episodes. But, 37 isn’t 30… ah, but 7 of the 37 clips are from the episode titled “11001001”, which is all about binary code!

This insight ended up being totally and completely wrong. I am perversely proud of this. Probably because I didn’t spend too much time running down this theory before I moved on to the last meta.

Pi Day

Clued Connections: Soo many different puzzle things. I joined a huge group, did some grinding, then left. I’m sure it was fun.

Compass and Straightedge : Didn’t spend much time on this. I had some teammates ask me if I had any ahas by looking at their data. In retrospect, I should’ve questioned whether they used the numbers as diameters or radii. And if you assume they’re diameters, you’re completely screwed. We ended up backsolving this, missing lots of bits of what seemed like it would’ve been a fun puzzle. Oh well.

Gone Guys : I joined up with four other teammates who had all the insights and knew how to extract, but needed grinding help from a logic-puzzle expert. I actually found the logic very fun, but I’m weird that way. At this moment I still don’t actually know what the extraction step was.

Moral Ambiguity : CSGP. I joined up a group of three other solvers, two in California who had one theory as to how the puzzle worked and wanted to solve on the spreadsheet, and one in Boston who had a different theory as to how the puzzle worked and wanted to solve on paper. I thought both theories could make sense. Fortunately, the puzzle came in four separate parts, so we were able to not get into each other’s way (actually the two Californians stayed on the same part). Eventually the Boston theory ended up busting, and poor Shelly left with a twinge of disgust. I ended up solving three of the parts (on paper, then typing into the spreadsheet). The extraction was pretty easy, fortunately.

Polyphony : Three poor folks had been working on this for 5+ hours. They had typed in a bunch of data, got the first two messages decoded and typed up in the spreadsheet… and that was all they had, abandoning the puzzle at 4pm on Saturday. I guess they didn’t know how to interpret it. Some folks looked at it at 2am, mentioned some theories on the chat, and then left. It was around 6am by the time I looked at it. One of the theories Jessen Yu mentioned on 2am seemed promising to me, so I figured I might as well try it. And it sort of worked. Then I found some errors on the spreadsheet data and it definitely worked. It required looking at three places at once, so I recruited a teammate to help me. Together we banged out the answer in 2 hours. I felt pretty satisfied at that.

True and False² : Shelley came to me and asked if I’d like to work on this with her. She had made some progress with Will (our other logic puzzle expert), but Will was tired and needed to sleep. I looked at the puzzle, confirmed her progress so far, and then realized that we had a philosophical dilemma. Should we treat this as a logic puzzle, and work on it together, making slow progress while trying to bifurcate carefully? Or treat it as a programming puzzle, and run a dictionary through it while seeing which ones made consistent solutions? The problem is, if the latter, Shelley couldn’t really contribute. Eventually we decided that there were too many bifurcation steps to do the former, and I got my programming skills ready. With some amusing hiccups (because of non-uniqueness), I solved the puzzle in one hour (including programming time).

Problems (Metas)

April Fool’s: When the team was stuck on this puzzle, I asked a “what if” question that turned out to be relevant: (“Jung vs guvf vfa’g n cher zrgn? Pbhyq gurer or bgure vasbezngvba ba gur cntrf gung jr arrq gb hfr?”) However, my question wasn’t picked up by anyone, not even me, so we didn’t solve it until hours later.

Halloween + Thanksgiving: Got the insight as to how the extraction was going to work (after 3 feeders). Other people did the work.

Arbor + Pi: Looked at it after all feeders had been identified. Got the insight and did the computing work to do the extraction. Turns out Brent Holman had a very close insight hours earlier which matched my first idea, but since I was a programmer I was able to tweak my code to try other things in a way that Brent wasn’t able to.

New Year’s + Holi: Our dedicated meta team didn’t recognize the shape (?!). I told them what it was. That’s about all I contributed.

New Year’s + Patriots: Last puzzle to be solved, and we were the only team to solve it. So hey, why don’t I write more details about what the last bits of solving it was like, since I was probably the most in the thick of this solve.

All the starting insights were already discovered (well, highly suspected) before I even started look at this. The thing is, even if you know what to do, getting the aha for the final extraction is hard because you don’t quite know what you’re looking for. For example, at some point we thought that there was Morse Code emerging at the top and bottom of the final image — it turns out that if you squint you can see TMENIST, which feels like it’s going to be something, but ends up being completely irrelevant. I had the unique advantage on the team at being (1) Well-experienced in gur Tnzr bs Yvsr to be able to have some intuition as to where to plop pixels; (2) Able to find a good app for quick keyboard-based stepping; (3) Having the biggest computer monitor onsite — I’m pretty sure I was the only person in a 300-mile radius to have a prototype “SPUD” — so this allowed other folks to collaboratively look at pixels with me.

When we were down to two normal puzzles and this meta left, that meant we had around 2–7 trigrams left to place (the actual number was 5, but we didn’t know that for sure). I was in the middle of writing a trigram-to-wordlist filter when Doug Zongker, working in California, finished his before I had even got the input working. So I abandoned that and worked on pixel/trigram guessing. Eric Prestemon, working in California, both felt that there was likely a trigram starting with T to clean up the mess on the right. I found TAC at about the same time as Eric — he loved it, I wasn’t as confident.

At this point there were a handful of teammates clustered around my screen. Most of them were not from our meta subteam and were regular teammates still trying to catch up and get a handle on how things were working, so I’d say that probably only me and Jasters (Jonathan McCue) were actually making progress on our Boston side. It was clear there were two directions of attack — either find more pixels to clean up the shape, or to look through Doug’s list and find answers that might be thematic (and still give appropriate trigrams).

Eventually one puzzle fell (forward-solve, since we did have 18 people working on it), and it was a nine-letter word… with the trigram TAC. Notably, it wasn’t in Doug’s output because (I think?) his wordlist wasn’t great. So I regretted a little bit at not having finished my filter. So, at this point, we know the last puzzle has an answer with 1–3 trigrams. At least 2, based on how messy the pixels were looking. I found some pixels to clean up the lower-middle, converted the trigram back and… it was QUE. Well, *that’s* not a common trigram. It felt like an eternity, but it probably only a few minutes until I realized hey, the last puzzle is a geography puzzle, the answer’s probably QUEBEC, let’s try that. I think Jonathan had the idea maybe at the same time. I immediately dived in, encoded the pixels, and voila, the answer grid looked really really nice. We already had a strong suspicion about the extraction method and were working on it for a few more minutes in excitement when it occured to us, hey, maybe we should actually *submit* QUEBEC and get it confirmed, so we did. Little did we know that the other room was less than a minute away from forward-solving QUEBEC. Anyway, the extraction happened, and we spent maybe another minute being confused, because the answer we got wasn’t a hilarious pun like the other metas, as far as we could tell — and it did seem awfully like yet another message telling us the next step. But we talked ourselves into submitting it, and that was the end of the metas.

Final Runaround

There was one “puzzle” at the start of the runaround. The chaos of having a team of 35 people trying to solve a puzzle that requires 10 physical artifacts to be manipulated, ideally in one location is pretty much what you’d expect it to be — led by people who were most assertive and knew what they were doing. I was one of those (maybe 5 such people). I provided a flat surface to collect the keys on and wrote the extracted letters on there.

For the rest of the runaround, I followed the team around and collected a bunch of data about the extracted letters just in case the puzzle was going to be something more complicated than “read the letters you are getting”. It was not.