The phenomenology of puzzlehunting

This is my write-up on MIT Mystery Hunt 2025. For context, see my write-ups from 2020, 2021, 2022, 2023, and 2024.

Why do people do puzzlehunts? It’s something I’ve covered lots of times, and my answer has always been: to have fun, to socialize, to be entertained, to be in awe… and, I guess, to solve puzzles.

I’ll admit, after my 2020 post, I haven’t written much about puzzles. But something about the 2025 Hunt put the experience of¹the experience of¹

Frustration

We can talk about whether people like doing difficult things or not, but that’s well-trodden ground, covered in plenty of game design tracts. The facts are: many people like finishing difficult goals, many people feel rewarded for doing so, and many people feel rewards proportional to the goals’ difficulties. Hoever, people only feel rewarded if they finish goals. Tension arises between designing a puzzle to be hard enough to feel challenging, and approachable after dedicated work. Something about flow probably gets mentioned too.

Further, not all kinds of difficulty are challenging. Solving two hundred cryptic crossword clues for a puzzle might be more tedious than fun. Challenge arises from difficulty via the locus of control that a solver feels. How much difficulty comes from not having a particular insight or skill, as opposed to artificial difficulty²artificial difficulty²

This year’s, I felt like I was stuck on the receiving end of two hundred cryptic crossword clues, several times. Sunday night was the most memorable.

I was in the ✈⁠✈⁠✈ Galactic Trendsetters ✈⁠✈⁠✈ headquarters. People were solving two big metas: some were working on The Killer, and some who were working on The Background Check metas. Our experience with The Killer has been documented in one of Brian’s Hunt posts, so I won’t dwell on it as much.

For The Background Check metas, I was probably stuck the hardest on The Grand Illusion. Like The Killer, it was a meta where we knew what to do, but struggled with doing it. In particular, (spoilers for The Grand Illusion)we had a map of Czechia with some points on it, and we needed to identify whether each point was within Czech borders, for each of four given dates. I couldn’t figure out how to find the data for this, short of reading a textbook on 20th-century European history. We solved it eventually after fudging some of our data. When I later read the solution, I learned there wasn’t much that I missed, other than failing to do research.³failing to do research.³

I think the vast majority of puzzles in this year’s Hunt didn’t have this problem, but negative experiences stick out. No one else is complaining about this meta. Cardinality did fine with it; how else did they finish hours ahead of everyone? Maybe it’s a skill issue on my part, then. If I wasn’t having fun, I should’ve done another puzzle, right?

But I didn’t want to work on another puzzle. The puzzles that contributed to finishing Hunt were The Killer, and the other Background Check metas. Plenty of people were working on each of them. I wanted to do the thing that was most useful, and solving The Grand Illusion seemed to be that thing. I could’ve asked for hints, but online hint responses were taking a while, and the Gala, where I could’ve asked for hints in-person, was a several-minute walk away.

And these are all personal problems. In team check-ins, the writing team always asks: Are you having fun? What can we do to make things more fun? Was I having fun? I didn’t know, at the time. And types of fun be damned, I’m not here to have fun, I’m here to sate my pride and competitiveness, and maybe I’m here to get hurt when it all ends.

Maybe. Probably not. Anyway.

If negative experiences stick out so much, then I, as a puzzlehunt author, might want to avoid solvers ever having negative experiences. Maybe I don’t want solvers to get hard stuck while solving puzzles. Maybe I don’t want solvers to be in the middle of a round, having solved a bunch of puzzles, but without a meta in sight. Setting aside the question of what that even means, this could lead to a guideline like: puzzles and rounds should be as small as they have to be, but no smaller.

Taken to the extreme, you get something like Galactic Puzzle Hunt 2024: hundreds of tiny puzzles⁴tiny puzzles⁴

Something about solving large (not necessarily difficult) puzzles makes them satisfying, beyond getting the insights needed to solve them. Something about solving metapuzzles and finishing rounds makes them satisfying, beyond solving an equal amount of puzzles. Can we delivered this satisfaction to solvers in a different way? Somehow, I doubt it, but it’d be interesting to see more attempts from others.

Further reading: Eric Berlin’s How Hard Is Your Hunt?

Missing out

The fear of missing out is a subtle aspect of Mystery Hunt,⁵aspect of Mystery Hunt,⁵

I’ve gotten used to the feeling at this point, but I was reminded of it this year. I’d invited several new solvers to Galactic, and was checking in⁶checking in⁶

I didn’t want to add to his FOMO, but there’s plenty of other ways it manifests. What if we finish the Hunt while I’m asleep? What if there’s an open puzzle I can one-shot? What if I’m spending too much time on puzzles and not enough time meeting up with all my friends?

While I feel a similar anxiety in other puzzlehunts, I think it’s most evident with Mystery Hunt. There’s a combination of factors that distinguish the Hunt:

• Size. How many person-hours does it take to finish a Mystery Hunt? To pull numbers out of thin air: if it takes 35 waking hours for a competitive team that averages 30 active solvers, that’s 1050 person-hours. If you have 150 puzzles, that averages to seven person-hours-per-puzzle, which feels about right. Let’s say it takes a thousand person-hours, give or take a few hundred.

  In contrast, a difficult online puzzlehunt, like GPH 2022, might take 120 person-hours. That’s an order of magnitude difference.

• Brevity. Mystery Hunt starts Friday noon, and (usually) ends by Sunday night. That’s two-and-a-half days long. This, combined with size, means you either hunt on a small team and work on most puzzles your team unlocks, but finish less of the Hunt, or you hunt on a larger team to see more of the Hunt but work on proportionally less puzzles.

  In contrast, the duration of a Galactic Puzzle Hunt is 10 days to span two weekends; several other online puzzlehunts also follow this schedule. A four-person team can spend 30 waking hours over 10 days, which would give everyone ample opportunity to work on everything.

• Physicality. The Hunt privileges those who attend in-person. Death and Mayhem stress this in the registration website. The physicality makes the Hunt especially ephemeral. There’s pressure to take in as much Mystery Hunt as you can, because the best experience will be during the Hunt, and anything that happens after is lesser. The shared presence makes Hunt a great time to meet up with friends, which also adds to the FOMO.

  In contrast, some puzzlehunts are more-or-less the same whether you experience it live or not. GPH 2023 and GPH 2024, for example, have post-hunt versions that are quite faithful to the during-hunt experience.⁷during-hunt experience.⁷

  Then again, a puzzlehunt is more than a set of puzzles, so the experience is nonetheless quite different; see my 2023 Hunt post for more on this difference.

Many puzzlehunts meet one factor, a handful meet two, but I’d bet only the Mystery Hunt meets all three. The tension comes from the tradeoff: small team or large team, see friends or go fast, explore or exploit.

How much choice do huntrunners have in this regard? The duration of Hunt is fixed and unlikely to change. The physicality of Hunt is also fixed: the endgame of a Hunt is an on-campus runaround⁸on-campus runaround⁸

If these global constraints are fixed, are there local mechanics that puzzlehunt authors can use to make people feel like they’re missing out on less? Some puzzlehunts unlock more puzzles automatically as time progresses, which could lessen the of, say, missing out on that one text adventure puzzle. In the 2021 Hunt, videos for the main story beats were available to everyone after a period of time, giving everyone a chance to experience the full story.

This year, D&M gave teams the option to choose which puzzle, out of a small set of options, to unlock next, given a short description of the puzzle. Like automatically unlocking puzzles over time, choosing which puzzle to unlock next can reduce the fear of missing particular puzzles. But unlike that, this also grants teams more agency: they can choose to unlock more puzzles, if they wanted to. And this mechanic has been discussed more elsewhere, as I’ll link in the further reading, so let’s move on.

Further reading: Dan Katz’s Unlocker’s Remorse: The Risks and Rewards of Choice as a Puzzlehunt Mechanism, and Drew Fisher’s Choose Your Own Adventure in the MIT Mystery Hunt.

Burden of effort

What exerting effort feels like is hard to describe in full, but it’s safe to say some people dislike expending lots of effort. Let’s discuss the distribution of effort, then. How is effort distributed between preparing for the Hunt and when it’s happening? How is effort distributed between people who write the Hunt and people who solve it?

Many of my thoughts about this have been influenced by working with the MIT Assassins’ Guild, a student group that runs live-action roleplaying games. When people discuss game design within the Guild, they’d sometimes draw the distinction between write-time effort and runtime effort. Common advice goes something like: beware the mechanic that’s quick to write but time-consuming to run.

An extreme example would be a LARP where every combat has to be adjudicated by a GM. There’s only a handful of GMs compared to players, so this doesn’t scale. The Hunt equivalent would be every single puzzle solving to a task instruction like SEND US A PICTURE, which would need a lot of submission review.⁹a lot of submission review.⁹

A similar contrast lies in the division between writer effort and solver effort. Suppose, in a LARP, that a GM wanted to give a player some item. One option is for the GM to find the player and hand them the item; this is writer effort. Or, the player can find the GM and get the item from them; or, the GM can drop off the item at a central place for the player to pick up; both have more player effort.

This year’s Hunt HQ was themed as the Gala, which had the vibes of a hotel lobby, at the R&D Commons in the fourth floor of Stata. D&M had an open-HQ setup, and teams could go to the Gala to pick up physical puzzles, ask for hints in-person, or interact with the writing team. The Gala was praised as a great way to interact with other solvers, and it certainly produced many good stories.

In the past few years, the Hunt HQ was the Bush Room, in 10-150. Given the spaces that I’m guessing teams hunt from, the Gala was on average farther.¹⁰on average farther.¹⁰

Puzzle pickup is one way that effort is offloaded from writers to solvers, but there’s other ways. Consider post-meta interactions. A common way to deliver story in a puzzlehunt is to drip-feed bits of the story after each meta solve.¹¹after each meta solve.¹¹

Consider Mystery Hunt 2021, which I wrote as part of Galactic. Meta solves were rewarded in two ways:

• Meta answers were used in post-meta interactions. We’d schedule a Zoom call with teams, and we played some small games with them.

• After some number of meta solves, we’d schedule a Zoom call between some cast members and the team, and the cast would act out a scene on the call.

Both ways involved runtime effort. Eventually, we skipped running post-meta interactions, and rather than scheduling cast meetings, we released pre-recorded videos instead.

Contrast this with this year’s Hunt. Story unlocked by meta solves were delivered in two ways:

• Answers to earlier metas led to dialogue tree cutscenes. I believe these were first introduced in the 2023 Hunt; see (spoilers for MATE's META)the MATE’s META cutscene for an example. These interactions happened in-browser, and the whole team voted on which dialogue options to pick. We used a similar cutscene format for GPH 2023 and 2024. I’m glad to see the 2025 Hunt use them, and with full voice acting too.

• Answers to later metas led to Gala interactions. The huntrunners scheduled a time with the solvers to meet at the Gala, or the solvers would look for someone specific at the Gala. The cast member would then take them to a separate area, where they’d play a small game and act out a small scene.

Cutscenes are a write-time effort. Gala interactions are run-time effort, but they involve more solver effort. In particular, teams had to visit the cast members in the Gala; this contrasts with cast members visiting the team, like what we did in the 2021 Hunt.

Two more questions I’d like to discuss, but don’t have the space to:

• Interaction puzzles, like the traditional scavenger hunt, or puzzles like this year’s Control Room or In Communicado Tonight. These involve scheduling variable-length timeslots. How do writers do this in a way that splits up effort appropriately?

• The endgame can take up to an hour and need up to a dozen people to staff, making it a drain on runtime resources. But the endgame serves as the Hunt’s climax, and writers often make it their goal for ten or more teams to reach endgame. How can a writing team deliver an endgame experience to everyone who finishes?

Now, why would a writing team want to shift runtime effort to the solvers? Sure, there’s a self-serving part to it. On the flipside, if the writing team does less things like scheduling interactions and walking between classrooms, they can do more things like hint-giving or team check-ins. And there was a lot of hint-giving, if you’re considering the hints given in the Gala.

I’ll close with a weird perspective on hints. We can think of hints as moving effort from write-time to run-time. Consider how we call a testsolve clean if it requires the testsolvers no hints to finish. For GPH, and I’d imagine many other puzzlehunts, a puzzle requires at least one clean testsolve before being considered done. Extra-important puzzles, like metas, might need more.

Why testsolve puzzles? The more clean testsolves a puzzle has, the likelier it’ll be that solvers also have a clean solve. So we can think of the testsolve–revision cycle, which is write-time effort, as something that minimizes hint-giving, which is run-time effort. Conversely, if we’re willing to offer more hints¹²willing to offer more hints¹²

Further reading: the section “In-game Workload” from Philip Tan’s thesis¹³Philip Tan’s thesis¹³

Conclusion

As I mentioned in my 2023 Hunt post: “we study media so we can apply (or avoid) what we’ve learned.” We criticize and analyze what’s been done in the past, in the hopes of making different things.

In this respect, Mystery Hunt 2025 is an excellent sophomore Hunt from Death and Mayhem. They gave solvers choices while avoiding the problems with their debut, they capitalized on the Hunt’s unique physicality, and they ran the whole thing without burning out.

In this respect, I’m also left with questions about what a puzzlehunt could feel like, inspired by studying this year’s Hunt:

• What does a puzzlehunt that minimized solver frustration look like? What about maximizing it? What about solver agency?

• What does a puzzlehunt that minimized the feeling of missing out look like? What about maximizing it? What about size, brevity, or physicality?

• What does a puzzlehunt that minimized the writing-to-runtime effort ratio look like? What about maximizing it? What about the writer-to-solver effort ratio?

• Can you play with these questions while still delivering a fun experience?

The design space for puzzlehunts is huge. I love how recent hunts have continued exploring this space, and I’d love it if hunts continued to do so. I’ll see you all at the next puzzlehunt I help write.

1. ¹

   Not the puzzles, not exactly, but the experience of working on them. Hence the title of this blog post, phenomenology, because we’re studying experience.

2. ²

   I’m reminded of a conversation the other night, where I was describing an idle game I was playing. The other person said it sounded a lot like work. I replied saying that work was different, because the performance of the company depends on innumerable factors I can’t control, but my performance in a video game was much more in my hands.

3. ³

   But maybe the point of the puzzle was to do this research—in which case, it wasn’t the puzzle for me.

4. ⁴

   Note that tiny isn’t the same as easy; lots of people can tell you about how hard AВΝС was. But making puzzles small minimizes the chance that solvers make partial progress before getting stuck. This does make puzzles more hit-or-miss.

5. ⁵

   It’s called the MIT Mystery Hunt, though, and missing out is also everpresent in the MIT undergraduate experience. There’s a huge variety of cool things happening all the time, and you only have a limited amount of attention.

6. ⁶

   This is not how this conversation went, but it’s close enough.

7. ⁷

   Then again, a puzzlehunt is more than a set of puzzles, so the experience is nonetheless quite different; see my 2023 Hunt post for more on this difference.

8. ⁸

   Exceptions are 2021 and 2022, which happened remotely. Still, both the 2021 and 2022 Hunts involved finding a coin on-campus.

9. ⁹

   Though see New You City from Mystery Hunt 2022.

10. ¹⁰

    The choice of the R&D Commons as the Gala area is partly logistical necessity. The Bush Room did not have a great layout for teams to walk in and out of, and it was also unavailable for pre-Hunt setup. The other option was probably Lobby 13, which is about as central as the Bush Room. It has worse acoustics and layout, and I don’t remember whether it was available. Such were the tradeoffs involved.

11. ¹¹

    Why reward meta solves like this? Diegetically, metas represent some problem, and the answer to a meta is a solution to that problem. These meta answers are also usually punny, and so surprising in their own right. It makes sense, then, to support these emotional moments with some story progress.

12. ¹²

    Though note there are other cost to giving hints, other than the run-time effort, which might make us not want willing. For example, solvers might find a puzzle less fun if they needed to use a hint to solve it.

13. ¹³

    The whole thesis is also a good read, and has lots of relevant discussion for the puzzlehunt writer.