Press enter to skip to main content.
Ask Us Anything Hunt Stats Finishers Big Graph Speedruns

Wrap-Up

Wow, what a year! Congratulations to the 294(!!) teams that finished the hunt, and to the 811 teams that solved at least one puzzle! The first three teams to defeat Kero and finish the hunt were:

  1. we left it too late to come up with something witty for the leaderboard ):
  2. idk
  3. And now, I said, let me show in a figure how far our nature is enlightened or unenlightened: --Behold! human beings living in a GalactiCosmiCave, which has a mouth open towards the light and reaching

Among the fastest solvers, this was the most closely contested Galactic Puzzle Hunt in history, with many teams in contention for the top spots.

We also had a speedrun leaderboard, and were very impressed to see competing teams sail past our internal fastest total time of 29m29.054s. Congrats to those teams, who all speedran our hunt in under 20 minutes:

This year, Galactic Puzzle Hunt had a significantly different format than in the past, with a smaller size and about half of the puzzles being based on the GalactiCardCaptors game. Creating GalactiCardCaptors was a wild journey for the writing team, and in this wrap-up we would like to share how it came into creation, to reflect on what lessons we learned, and to muse on some puzzlehunt philosophy.

If you want to jump ahead, you can find more fun stuff at the end!

Hunt summary

At the very entrance of the GalactiCosmiCave, you find a treasure chest, surely filled with great riches. But inside the chest is Kero, a magic carp! Kero informs you that opening the chest has unleashed CHAOS throughout the cave! ＼(º □ º l|l)/ Fortunately, Kero also gives you the uwuand, a magical wand that converts cave dwellers into cards friends that you can use to battle! Using the power of the uwuand, by solving puzzles and playing GalactiCardCaptors, you befriend the 12 legendary creatures living in the cave.

But, all is not well; for every legendary that you befriend, Kero gains a body part and grows stronger. After befriending all 12 legendaries, Kero turns on you! With the legendaries united, you defeat Kero in a final battle. You are given one last choice: seal Kero back inside the treasure chest, or invite him to join you for some pie?

Look carefully and you’ll see what the first team chose:

✈
✈
✈
we left it too late to come up with something witty for the leaderboard ):
Legendary
.
- bryan
- brian
- pie
- navi
- kevin
- eggy
Everyone loves Pie!
∞
∞
(you can check out the Victory page too!)

If you want to see the plot in more detail, go watch all of the cutscenes!

Final faction scores

Near midway or thereabouts through the hunt, a team could choose to align itself with one of the five factions dwelling in the cave, and therewise earn a smidgen of points during battles.

Golly, we haven’t seen a closer race in millennia! I daresay we Dinos were on the brink of settling for a humble bronze, but lo! That dash of vim, that final surge! Methinks we right put those Cows in their place, did we not? (Granted, we’ve since reconciled our differences, but who doesn’t relish a good-natured tussle?)

Hats off to The Coraline Schism, whose efforts furnished a staggering >50% of our final tally. Yet, let us not forget all our other Dino supporters, for without your stalwart backing, our triumph would’ve been naught but a dream! Huzzah, huzzah, to one and all!

Writing the hunt

Goals

Our main goals when we kicked off hunt writing were:

Timeline

Theme

From our past experience writing GPH and MIT Mystery Hunt, during the theme proposal phase we have seen several “arms races” between different theme ideas, trying to be as fully developed as possible to maximize the chance of being selected. While this might make the hunt-writing later a bit easier, it risks wasting valuable effort that went into other unselected themes (even though they might be re-proposed later).

Therefore, this year we decided to have two phases of theme proposal:

We received fourteen theme pitches in the first phase, and the winning pitch was Galactic Card Patrol:

Deployed across the galaxy, teams solve puzzles to collect Galactic Cards, manifestations of ancient alien technology, which they use to unlock secrets and overcome challenges. These cards have unique abilities, and can be combined for unexpected and powerful synergies.

In the second phase, there were seven proposals, including two “partial” ones that were meant to be integrated into other proposals if possible (e.g. as one round or an additional mechanic). When voting, we considered how exciting the theme was, how willing people were to work on the theme, as well as strengths and potential risks of the theme. The winning proposal, GalactiCardCaptors, was centered around the premise that “each puzzle is a bespoke card battle with an enemy, and defeating it captures them into a card”. While this premise was somewhat relaxed during the writing process, much of the original vision made it into the hunt you played!

(Click to read the original theme proposal.)

As we had hoped, some aspects of the unselected proposals from the second phase also managed to make their way into the hunt in some form:

Answerables

As early as announcing the winning theme, we decided to include answerables, traditional puzzlehunt puzzles with answers, in our hunt as well. (This may seem obvious to experienced puzzlehunters – keep in mind our theme proposal was comfortable leaving answerables behind entirely!) More than just being familiar staples of hunts, we wanted answerables to diversify the experience of the hunt (allowing both writers and solvers who are not so into card games to contribute), as well as possibly provide some additional benefit to the card game. We intentionally asked writers for answerables to be short to not draw too much attention away from the main theme.

Ideas for answerable rewards included card upgrades (à la Slay the Spire), additional copies of existing cards, card cosmetics, and awarding “booster packs” of new cards – the last of which ended up in the final game!

When possible, we tried our best to match thematic card rewards to answerables (as well as non-legendary battles): the two best pairings are getting Quail Aboard Fiery Steeds with Limericks and getting Eager Beever after beeBay Fulfillment Center.

We have more thoughts about answerables in our Reflections section.

Card game design

As soon as the theme was finalized, we started drafting, playtesting, and iterating on designs for what would eventually become GalactiCardCaptors.

The first design of the game (from the theme proposal) was inspired by Inscryption/Air Land Sea/Marvel Snap, with players taking simultaneous turns, opposing cards attacking each other, and dealing damage to the player’s “life points”. This design then evolved to a larger, 6x5 grid with different terrains that cards automatically moved down – and the win condition changed to controlling three or more columns.

Concept art for GCC “rock paper scissors chess”. This was the first design to be prototyped and playable.

We experimented with many additional features, like cards having “sets”/”suits”, and additional “types”/”elements”. Inspired by Radlands, a tabletop card game, we drafted a new design centered around defending Bases, cards which produce Energy for other cards to take actions.

This latest design also added six colors to the game, with cards from each color having different playstyles. Sound familiar? This is also when “protection” – placing cards in front of other cards to defend them – was added.

When we playtested the GCC “Radlands edition” in Tabletop Simulator, we found that the emergent gameplay of the Energy mechanic (later renamed to Gems, which finally became Food) created deep strategic choices: should we attack with a creature this turn, or create food to play stronger creatures? How strong of a board presence do we need to “ramp up” to before we can attack? Should we defensively protect our Camps first, or attack aggressively? Finding these classic card game tradeoffs in this design really solidified this ruleset, and we found ourselves having a lot of fun playing matches with basic cards against each other.

Having finalized our base game, we went to work getting the game prototyped and playable so that we could start designing cards, and most importantly, designing card battle puzzles. Once we had the basics of the game implemented, we hosted a writing retreat for everyone to learn the rules of the game, play in a big tournament, and start iterating on battle puzzles.

Battles

We have lots of interesting stories about writing our battles: see our Author’s Notes in the solutions for each battle!

Early on, we divided battles internally into what we called “just battles” and “puzzle battles”, the latter of which would involve more enigmatic insights, like Coloring and Slime. This distinction ended up not really mattering both publicly and internally, especially since there isn’t a clear line between the two. Once the Kero fight was designed, we decided to communicate the “legendary battle” (👑) and “non-legendary battle” (⚔️) distinction instead, since it was more important thematically and structurally to know which battles were key to solve. All our battles are puzzles! 😙

Testsolving

We had three “big testsolves” (BTSes) starting from August to October. The first (Small BTS, or SBTS) was internal and tested about half the battles (including Mastery Tree and the Kero fight). The following Medium BTS (MBTS) and Big BTS (BBTS) tested all the battles and most/all of the answerables. Our hunt really came together thanks to our BTSes, which helped give early feedback and iteration cycles on our hunt flow, puzzle & game design, and user experience. Thanks so much to our BTS testsolvers for making our hunt as polished and cohesive as it is.

Reflections

Overall

We knew that the card game could be polarizing. There were a variety of considerations that went into choosing it anyway, but ultimately, we wanted a cohesive whole-hunt story/experience and this theme generated the most interest on the writing team. This went more or less as expected: we received feedback that some solvers who enjoyed past GPHs didn’t enjoy this one at all, but also many solvers who said this was their favorite recent hunt, or even favorite hunt of all time. Of particular interest, we got feedback from multiple solvers who didn’t expect to enjoy the hunt at the beginning, but ended up loving it by the end, which we are very happy about. That said, we appreciate all the solvers who tried the hunt and didn’t like it, and we hope that you will find future GPHs to your liking.

What is a puzzlehunt?

Is this a puzzlehunt? This question has been the topic of much discussion, both internally and externally. It turns out that this is not just semantic nitpicking, but an important aspect of how we talk about and advertise hunts.

While there isn’t a definite answer to this question, the best alternative we could think of to describe our hunt was a “puzzle game”, and we do think, at least, that our hunt is closer to a puzzlehunt than it is to a puzzle game. (We also considered something more vague like a “puzzle experience”...). The most important reason we decided to call GCC a puzzlehunt was to push the established boundaries of hunts rather than defining ourselves as outside that boundary.

Even if we excluded the 18 short answerables and the mastery tree meta, we think that many of the battles’ primary mechanics were closer to something you’d see in a hunt puzzle than in any puzzle or strategy game, in that they are novel mechanics that appear in only that puzzle and often aren’t explained; figuring out what they do is more important to the solve experience than figuring out their consequences.

Examples include deciphering the words in Dargle, determining the symbols in Jabberwock, understanding the rules in Moonick, and understanding the slime splits in Slime. Additionally, some puzzles like beeBay and Spelling Bee require minimal card game strategy and mostly feel like self-contained interactive puzzles.

On the other hand, there were also several puzzles that required a significant amount of strategy to solve (e.g., defeating Slime).

A highly subjective table might look like:

\ Game insights

Nongame insights

low medium high
low Angry Boarry Farmer, Gnutmeg, Infinitely Many Hogs (and Logs) bb b, Asteroid, The Swarm, Kero, Spirit of the Vines
medium Spelling Bee Dargle
high Moonick, beeBay Othello, Jabberwock, Coloring, Miss Yu, Mister Penny Slime, Blancmange

At an even higher level, the way we envisioned most solvers would participate in our event — by forming teams and collaboratively doing puzzles at the announced time the hunt was running — is much closer to a puzzlehunt than a puzzle game.

We should mention Jack Lance (the real Captain Pi)’s Root Three Riddle Search, whose release coincided with our theme proposal deadline. While we weren’t directly inspired, we think the RT3 puzzlehunt reflects similar thoughts we have about pushing the boundaries of traditional answerables and puzzlehunt structure.

Pre-hunt information

When we first announced the hunt, we were very conservative about what information we released, since we too had very little idea what to expect from the hunt at the time. It wasn’t until receiving feedback from BBTS when we knew that many solvers would appreciate more pre-hunt information to set appropriate expectations about the hunt.

Even then, just how much more information we wanted to release was a subject of much internal debate. On one hand, we wanted to preserve the surprise and the excitement of speculation on the theme – which many of us felt was an integral part of the hunt experience – as well as avoid creating preconceptions about the hunt and what solving it would be like. On the other hand, expectations can have a very large impact on how a solver interacts with different aspects of the hunt’s mechanics, and, ultimately, their solve experience – and, in our case, the precedence of prior GPH shenanigans might not have been enough to prepare solvers for just how unusual this hunt would be.

We ended up adding an FAQ item revealing that many puzzles would feel more like “puzzle games” than puzzles in previous GPHs, but that there would still be traditional puzzles. We tried to make it stand out by adding an alert to the top of the About page, and encouraged solvers to read that page in an email we sent out a week before the hunt. We hoped that this would strike a good balance by widening solvers’ expectations without giving too much away.

We had also considered releasing some information about hunt difficulty, but decided against doing so in the end. We knew from prior experience that hunt difficulty was very hard to predict, and were wary about making promises about it before the hunt. This was particularly the case this year, as we expected that the card game aspect would create a large amount of variance in difficulty for teams.

Our approach to managing solver expectations worked for some but not others. Many solvers were still caught off-guard by the hunt’s format, and some have expressed that it would have helped to know more about the role that battles and answerables were intended to play in the hunt. As hunts continue to experiment with new and interesting ways to break the mold of the traditional hunt format, we hope that this hunt would serve as a useful precedent to help future hunts set solvers up to get the most out of their hunt experience.

Tutorial

Designing the tutorial was challenging. We needed to communicate a large amount of information to solvers in a way that was engaging, easy to absorb, and accessible to solvers with all kinds of backgrounds, all without feeling too long or handhold-y. To this end, we subjected the tutorial to more playtests, revisions and tech work than any other puzzle in the hunt.

One of the key design choices we made was to introduce game mechanics through small puzzly challenges as much as possible. We hoped that this would not only make the tutorial more engaging for our target audience, but also help solvers absorb information better by immediately applying what they’ve learned. This philosophy informed the design of stages 3 and 4.

We also wanted to ease solvers into the more strategic, open-ended thinking required in many battles. This led to the creation of stage 5, and the intentional placement of Angry Boarry Farmer right at the start of the hunt.

The tutorial we ultimately released had its flaws. Teams found it harder than we intended, with a large percentage (~5%) of teams spending a lot of time on stage 5 early in the hunt. In response, we issued an erratum clarifying that the enemy battler was not food-limited, along with some tips for that stage.

Keepsakes

Part of teasing the hunt theme involved sending GCC keepsakes to early donors from our 🪙 donation page 🪙 when we announced the hunt. We were hoping this would subtly clue our theme, that GCC stands for something other than GalactiCosmiCave, and that cards would be involved.

We were hoping that keepsakes could stoke public discussion of theme predictions, which we have a lot of fun reading. Unfortunately, this didn’t really happen much. Also, many US recipients had their keepsakes lost in the mail, possibly due to border issues. We’re very sorry about this! If you claimed a keepsake in August, never received it, and haven’t contacted us about it, please let us know ASAP at contact@galacticpuzzlehunt.com.

That said, we later learned that at least one team correctly guessed (in private) the Cardcaptor Sakura reference from their keepsake!

A lot of work and care went into designing the keepsakes, getting them printed, and mailing them out, but it was worth the effort to thank you for supporting our hunt! If you’d like to support our work, please see our 🪙 donation page 🪙. There is a limited number of additional keepsakes available.

The Future of GPH

Will there be a Galactic Puzzle Hunt 2024? Join our mailing list to be the first to know!

In the interim, you should check out some of these other puzzlehunts coming up:

Check the Puzzle Hunt Calendar for more. We’re excited to see what new ideas the community comes up with!

More reflections

This section goes into quite a bit more detail about the various aspects of the hunt-writing/designing process, written by the leads of different departments of our team.

Philosophy

Puzzlehunting is a hobby that involves many diverse interests. Despite the open-ended nature of puzzles without instructions, the structure of hunts have crystallized into a standard format: solving answerables, puzzles with answers, which are used as “feeders” in meta-answerables that act as “capstones” for progressing the hunt story and structure.

Puzzle answers are single words or short phrases, so answerables are designed around the solver trying to condense all the information into a single word or phrase.

This has led to a lot of institutional knowledge over time:

Sometimes, answers are also forced into interesting standalone puzzles that are fantastic on their own and have no need for an extraction outside of the hunt – from classic puzzles like crosswords and logic puzzles to activities like interactive games and scavenger hunts. This definition of a “hunt puzzle” is extremely narrow, despite the fact that we like to explain to newcomers that they can “be about anything”. It’s no wonder we have a hard time explaining puzzlehunts!

This standard format also offloads much of the effort of writing thematic and cohesive puzzles to metas themselves. While metas can place interesting constraints on their feeders, it’s often the case that a feeder answerable is free to use nearly any mechanic, as long as it solves to the assigned answer. By design, authors are free to write any answerable they want – at the expense of a less narratively relevant and thematic puzzle.

Because of these thoughts, one of the main goals of our hunt was for every puzzle to be important, thematic, worth engaging in, and part of the structure of the hunt. Solving each puzzle should be its own reward, and we wanted to explore other ways of tying together puzzles into a hunt package. As you can read from the winning theme document, the initial proposal was so far along the “anti-meta revolution” that it only described card battles, which were intended to be the full experience of hunt; the interaction between capturing new cards to use in other battles (and losing cards in special battles, like the “black hole”) was a cohesive and rich enough design space that metas (and answerables) didn’t feel necessary to tie together the puzzles.

In the end, the final Kero fight became more of a traditional metapuzzle (tying all the legendary cards together), and the Mastery Tree round was introduced to provide more variety to the hunt, so we leaned back a bit from the initial purist anti-meta vision. To those of you debating whether our hunt was a puzzlehunt at all – consider how far we were willing to go! In fact, many of our theme ideas this year involved novel ideas that would break established puzzlehunt patterns.

We hope the context change that GalactiCardCaptors provided gave beginner teams and the larger puzzle-curious community more equal footing with experienced teams. There is a point in our hunt when every experienced team says, “we finally unlocked real puzzles” – a value judgment that only those entrenched in the existing framework could have. We hope to see the puzzlehunt community grow and continue to innovate beyond answerable puzzles, and stay inclusive and open-minded of attempts to do so.

Puzzle design

The structure of the hunt this year brought with it many challenges for puzzle writing. Over the years, we’ve learned quite a lot about designing round structures, metapuzzles, and traditional hunt puzzles, but much of that knowledge was not very useful to us this year when we started working on card battle puzzles.

For a few months after deciding on the theme, we weren’t sure exactly what card battle puzzles should look like, and it was especially difficult to design them for a variety of reasons:

For these reasons, one of the top priorities at the beginning of the writing process was to get a few puzzles written as soon as possible, so that other writers could see what a card battle puzzle could look like, and to help solidify the process of writing card battle puzzles.

During this period, there were significant concerns that we would not be able to run the hunt as planned, because it was really quite tricky to get the first puzzles written. It took until early July to get the first card battle puzzles into a testable state, and (along with the significant amount of technical work required) was one of the reasons for this year’s hunt happening later than previous hunts.

Some of the first puzzles to be fleshed out and written were bb b, Dargle, and Blancmange (although they all had very different names at the time). Some trends that we noticed:

While we are happy with the set of card battle puzzles we came up with, a lot of us on the writing team learned a lot about designing these types of puzzles after the internal “big testsolve” events, and might be able to come up with a lot of new, interesting ideas if we were to design a new set of card battles now 😉.

The Mastery Tree and answerables

The answerable puzzles (i.e., the puzzles with traditional puzzlehunt answers) in this hunt were intended to complement the card battle puzzles, both for the writing team and for solvers. Some reasons for the inclusion of the round, from my perspective:

We targeted a lower level of difficulty for these puzzles, partially to accommodate newer and less involved authors, and also because we didn’t want to make the answerable puzzles blocking items for the card battles. For this reason also, solving answerables does contribute towards unlocking more puzzles, but not as much as solving card battles.

The first answerables in the Mastery Tree round only unlock after three card battles are solved. This was done primarily to encourage all solvers to give the card battle puzzles a try, as we felt that unlocking them simultaneously with the card battles would result in solvers who started with the card battles becoming “specialists”, whereas solvers who started on answerables would feel like they were too far behind on card battle knowledge to engage with it. This was something we had observed in the past with Puflantu in 2019’s hunt. We also did this to set expectations that card battles were not just a puzzle “gimmick” but the backbone of our hunt.

For some teams, having dedicated sub-teams working on the card battles and answerables might be ideal, as some solvers feel strongly about what types of puzzles they prefer to work on. However, we did get a lot of feedback from solvers who initially felt like the card game wouldn’t be their type of thing, but ended up really enjoying it.

Hunt structure

Since our card battles did not have traditional answers, there was some discussion early on about what sort of capstone puzzle we should have for the hunt. There were some pretty adventurous ideas, like having a simultaneous battle at a fixed time, with all hunt participants chipping away at a “raid boss” encounter. The “final battle” and legendary cards felt like a good way to bring in the fun of progressing towards a final puzzle in a way that made more sense for the card battle system.

One interesting aspect of it was that it required solvers to complete a much greater fraction of the puzzles than in a typical hunt, in which metapuzzles can often be unlocked and solved with far fewer than all of the “feeder” puzzles. We considered adding some kind of leniency, like giving solvers the last legendary card if they were only missing one, but decided against it as it did not fit well with our hunt’s theme and mechanics. Moreover, we wanted to have the flexibility for legendary cards to include spoilers for their source puzzle, or rely on having solved the source puzzle to be understood correctly. Instead, we opted to rely on the hint system to get solvers past puzzles they were struggling with. By moving the initial hint release to 12 hours into the hunt instead of the usual 24, and targeting the puzzles easier than our usual fare, we hoped that the requirement to solve all legendary battles would not become a frustrating wall for solvers.

Testsolving

We found that this hunt was more difficult than previous hunts to testsolve adequately. There were numerous enthusiastic solvers on our team who tested essentially all of the puzzles, but we felt that they would likely find the puzzles to be fairly straightforward even if solvers in actual hunt conditions would find them quite difficult, due to our testers being much more familiar with the battle system. On the other hand, it seemed like testers trying out their first card battle puzzle might find it a little more difficult than we’d expect solvers in the actual hunt to (especially for puzzles towards the end of the unlock order).

For this reason, we placed a lot of emphasis on three “big testsolve” events for tuning the puzzles and hunt structure (named SBTS, MBTS, and BBTS: “small big testsolve”, “medium big testsolve”, and “big big testsolve” respectively). We ran SBTS with about half the hunt written (including both the Mastery Tree and the Kero battle), primarily with team members who hadn’t been involved with the card battles up until that point, and then ran MBTS and BBTS with completely unspoiled solvers, with hunt writing mostly complete. Even more than for a usual hunt, these events proved particularly important in tuning puzzles, and seeing how teams might interact with the unlock structure and hunt flow.

A Story about Story

Way back in the days of yore (a time that probably only the dinos remember), a game prototype was made with colored gems as resources. The story began in a cave, where you found a crystal. You would explore the world, find help along the way, and eventually go to space to fight for the fate of the worlds! (Evil Entities trying to farm planets for the power of gems)

Then, came the factions. We hoped that each kind of faction might lend itself to a kind of puzzle. (e.g. the bees like spelling, the lobsters liked memes, etc.)

Come June, new blood was injected into the story. To the forefront came two ideas: cake, and a cute companion named Cerberus (inspired by Cardcaptor Sakura). Cerberus, we thought, could also be a play on the guard of the underworld, and perhaps Sakura was the Queen of the Underworld.

We wanted to stitch together enough of a story to spark ideas for puzzles and cards during writing retreat, so we made cake:

One concept we played around with was for dragons to get more powerful as they got longer, which brought us the cards o and re (who then moved on to become cows later on in their lives).

But more importantly, you ask, where are the boars?!?!

Perceptive of you to ask! During that writing retreat, we playtested what cards we had (made by Charles) in a PvP tournament. The Boar was OP, defeated many an opponent, and wooed many a heart. The retreat ended with some Gartic Phone activities in which some boars on the floors became hogs on logs, and it just became very clear that we needed more boars. MOAR! So, they replaced the dryads, and made boarries. And they got their own Discord channel. Twice.

We were also left with a couple of tough questions. Do the factions live in harmony or are they at odds with one another? In the narrative, are you playing card battles, or are the card battles actual battles? (In a funny turn of communication, proponents of both options called this “playing it straight.”)

So naturally, having finished with writing retreat, we set aside the questions, and for two months, there was little more movement on the story (this is an important ingredient of puzzlehunts — letting people and creation rest) until there was another push because of the upcoming SBTS (small big testsolve) (another important ingredient of puzzlehunts — deadline driven development). At this point, the Mastery Tree and Final Battle had been spun into existence.

The story team looked at Captain Pi and thought, we need pie.

And thus, cake became pie. For pie, one must have butter, and thus the cows changed their industry to make black butter and whipped cream (conveniently, B and W, not that the other factions have the letter of their color and resource match).

During this riffing session, we also thought, wouldn’t it be cool if Kero took on the traits of the factions, much like the Chinese Dragon is made of different animals? “Oh!” said another, “He’s Kero Dos!” And thus, it came to be that the original Kero had to be a magical carp. (Yes, it’s a play on Gyarados – who we learned later is based on a related myth!)

This led to some other conundrums.

What does the gnome provide the dragon? We must have antlers. And thus the dryads made a resurgence and replaced the gnomes, who took their silent G’s and found their way to the Gnutmeg Tree.

Why is it that Kero becomes a dragon, one of the factions? And thus, the dragon faction became the dinos. (Which, ironically, makes them actually the youngest of the factions.)

So were the factions set, and the Final Battle rethemed to relieve Kero Dos of his ill-gotten powers.

During this super-serious riffing session, it was joked that uwu could be Underwater Usurper, and because we are a very serious team, this became reality. Adding Rock Lobster was only natural at this point.

With all the story pieces set, we wrote up some example dialogue with the intention of not branching, but in the end, we couldn’t decide if the better ending was to forgive Kero and invite him to eat pie with us or lock him up in a box and start the cycle all over again. So then … the web team … just made that happen. Despite them already being underwater (hah).

The puzzles were written, and to each battle, a faction was assigned. But the factions needed homes! It was only natural that the map be split into five sections, one for each faction.

But alas, this was not to be. There were more important factors to consider when arranging battles, such as difficulty, variety, and card rewards. And so the battles were arranged without due concern for factions. The draft order looked like a mess at first, but then, a closer look revealed a most intriguing pattern. The first five battles were one of each faction, and beyond them, the battles were mostly neatly split across faction lines! You just had to ignore the cows, which were everywhere.

And that was when the truth dawned on us. Cows never belonged in one place! They were nomadic. Or, perhaps, they were just everywhere – a sprawling industry that touched every corner of the cave, a massive tax collection apparatus! And then, later, when we realized that we had no non-legendary cow battles, our eyes were opened a second time. The cows were powerful, arrogant creatures that roamed wherever they pleased.

Art

All the cards.
A real card from a real TCG that served as an inspiration for us to accept a wide variety of art styles.

We made a LOT of art. Like, almost everyone on the team did some art or another. See our art credits!

A couple decisions we made earlier on were:

Concept and “beta” art

Original prototype for Kero’s transformation: https://scratch.mit.edu/projects/900412088/

Original prototype for Kero Dos: https://scratch.mit.edu/projects/907519553/

Web

Setup

The interactivity server relied on typia for validation. We used JWTs to authenticate clients to the Node servers based on accounts from gph-site. We did not rely on any frameworks for the game engine itself.

Diagrams of the prod stack and various dev environments.

The advantage of this setup was performance. After our experiences with Django Channels in previous years, we were particularly wary of using Django Channels in any large capacity. (In fact, we still don’t have a definitive answer to what went wrong last year, though we have a few more ideas from our experience with load testing this year.)

The disadvantage is that restarting the interactivity server would have been very disruptive. This isn’t really a problem with this setup per se, since getting blocked out of an interactive puzzle for a few minutes would have been an unpleasant experience regardless of how it was set up. Given that most of the hunt was going to depend on interactivity, we chose to focus on making a server restart as unlikely as possible, instead of building in a graceful shutdown mechanism. (We did end up making a graceful shutdown protocol as a contingency, but it required users to perform a manual refresh.) Apart from doing lots of load testing and manual runthroughs, we were prepared to, and did, make heavy use of the Node debugger’s REPL to monkeypatch code on the fly in response to serious bugs or errata. We did not need to restart the server at all throughout the entire hunt’s runtime.

One thing we enjoyed a lot about the new infrastructure (but really, the highly interactive nature of the hunt itself) was the ability to spectate teams’ battles directly. We also enjoyed upgrades to the big board, the page that displays a summary of each team’s progress through the hunt. The new big board updated in real time without adding significantly to server load. It also displayed which battles each team was actively engaged in.

Load engineering

We made load engineering a priority for this hunt, and were glad to see that it succeeded. For context, our past hunts have been plagued with load problems, particularly last year when the site crashed as often as multiple times an hour. This caused a lot of solver frustration. We wanted to avoid this in this hunt to the best of our ability, particularly given how much of the hunt depended on interactivity.

Since we barely had any load engineering experience as a team, we made sure to set time aside for it this year. We started load testing in earnest just after MBTS, 1.5 months before the hunt date. This was actually a few weeks later than planned, but, fortunately, it worked out for us in the end.

Here are some highlights:

UI

For past GPHs, nearly everything frontend-related was built with Django templates, with no other frontend framework, because the UI was mostly simple. (It also helped that we had a great base, gph-site, to build on.) This hunt, in contrast, needed a lot of UI work, which is why we chose to add a React client early on. A lot of the inspiration for the battle and deckbuilding interface comes from existing digital card games, including MTG Arena, Hearthstone, Inscryption, Marvel Snap, and Slay the Spire. Interface In Game was a great reference for how other games structured the UI.

Development timeline
Gallery

Fun stuff

Team shoutouts

More stats

Links

Fan art (send us more!)

From @RedBirdRabbit (original tweet):

From @tetaes001 (original tweet):

Funny answers

Trigram Hell
Petroglyph
Limericks
Make Your Own Star Battle
Now I Know
Gessner Mom
Console-ation Prize
Beige Flags
#vent
Paint By Numbers

:praytrick: submitted:

Mastery Tree

Some teams got close to the answer…

Some teams tried answering the question directly…

Gauss’s Sketches

if MATE has 1 MILLION fans i am one of them. if MATE has 1 THOUSAND FANS i am 1 of them. if MATE has 10 fans i am one. if MATE has 1 fan i am the fan. if MATE has no fans i am dead. submitted:

Fun stories

After the infamous writing retreat tournament where Boar (now Boarry Farmer) was extremely strong and popular, our card ideas channel was full of new Boar cards:

And thus, the Boar faction was boarn.

We can’t forget Boarbie, who fell into relative obscurity...

Much of the inspiration of Cows come from these legendairy videos by ProZD.

One cut mechanic from GCC is the ability to customize the Bases you start with in each battle. We imagined being able to upgrade them and even specialize for particular factions (like producing 2 Boarries instead of 1 Pie). Here are some faction-specific Base concepts that never made it into the game:

There were many polish features that were almost cut from hunt because they caused our web team stress right before deadlines:

We’re glad they made it in, we personally love them and we heard lots of positive feedback too!

WTFTadashi: I figured out Now I Know while driving my 4 year-old daughter home from pre-school. We had worked out that the words were shared between Baa Baa Black Sheep and Twinkle Twinkle. I was thinking about the puzzle in the car, and I kept singing the two songs and thinking about where the words overlapped. At one point, I was humming the tune, and my daughter started singing the A,B,C’s. That’s when it occurred to me that we had to also overlap the songs with the Alphabet Song. I suppose I should add her to the team roster now.

Gibbon Group managed to predict the final battle:

They also built Animal Shelter in Minecraft:

The Hunt Slayers appreciated the Slimes “a healthy amount”:

Eggplant Parms had a special strategy for Infinitely Many Hogs and Logs:

Credits

Chairs: Amon Ge, Yannick Yao

Art Leads: Chris Jones, DD Liu

Art Credits

Story: Kat Fang, DD Liu

Taskmaster: Jenna Himawan

Scribe: CJ Quines

Head Editors: Anderson Wang, Colin Lu

Editors: Charles Tam, Lewis Chen

Card Game Design: Charles Tam, Amon Ge

Onboarding Lead: Josh Alman

Email Lead: Alan Huang

Factchecking Lead: Jenna Himawan

Testsolving Lead: Kevin Li

Web Lead: Mark

Web: Alan Huang, Amon Ge, Andy Hauge, Brian Chen, Charles Tam, Chris Jones, CJ Quines, Colin Lu, DD Liu, Dominick Joo, Jenna Himawan, Jinna, Violet Xiao

Puzzle Writers: Amon Ge, Anderson Wang, Ben Yang, Brian Chen, CJ Quines, Charles Tam, Chris Jones, Colin Lu, Dai Yang, Jenna Himawan, Jinna, Josh Alman, Kat Fang, Kevin Li, Lewis Chen, Mark, Nathan Pinsker, Patrick Xia, Violet Xiao, Yannick Yao

Factcheckers: Anderson Wang, Andy Hauge, Charles Tam, Colin Lu, Dai Yang, Dominick Joo, Jenna Himawan, Jon Schneider, Max Murin, Sam Kim, Wayne Zhao, Yannick Yao

Postprodders: Amon Ge, Brian Chen, CJ Quines, Colin Lu, Jenna Himawan, Violet Xiao, Yannick Yao

Additional Testsolvers: Alwina, Andrew He, Bex Lin, Cameron Montag, Catherine Wu, Connor Tilley, Harrison Ho, Ian Osborn, Jacqui Fashimpaur, Katie Dunn, Leo Marchand, Leon Zhou, Lumia Neyo, Maddie Dawson, Rahul Sridhar, Steven Keyes, Tom Panenko, Vlad Firoiu