anderson's Puzzle Blog

Galactic Puzzle Hunt 2017 recap: the puzzles

leave a comment »

The Galactic Puzzle Hunt took place about two six weeks ago, and overall, I had a great time helping create and run it and I’m really proud of everyone who contributed. There’s quite a lot of things I want to talk about, so this post will be dedicated to going over some of the specific puzzles; there will be another post for organization and other notes. There will be lots of spoilers, so be warned.

Before I get into the details, here are some personal puzzle recommendations if you haven’t seen them yet and want to try them out before being spoiled:

Now, for some specific puzzle commentary:

Puzzle of the Day: I don’t really have too much to say about this; I think it’s a good easy puzzle and I’m glad so many teams solved it.

Angry Portals: Yeah, so the last page of this puzzle (“urlsarenotapuzzle.html”) turned out to be a huge red herring, oops. In testsolving, I suggested that we change it to “urlsarenotapuzzleandneitheristhispage.html” or something like that, which might have helped. However, even our intended message of URLs not being a puzzle kind of failed, because as several teams pointed out, the final page depicts a cake and “the cake is a lie”, which means that the URL name might also be a lie! We 100% did not see this coming, but in retrospect it makes a lot of sense.

Other than that, I think this puzzle is really creative, and just solving the levels is a fun puzzle in its own right. One thing that isn’t in the author’s notes but really should be mentioned is that DD spent a lot of time revising the puzzle based on our suggestions: in particular, the game originally didn’t have an undo function, which made several levels much more frustrating, especially the last one. So I just want to give huge props to her for all the time she put into improving this puzzle.

Learn to Play: I also really like the creativity of this puzzle. Also, one of my favorite anecdotes relates to this puzzle, so I’ll share it here (from ET Phone In Answer): “Here’s how our team member Anders figured out Learn to Play. He looked at the exact div widths in the source code and figured out sequences of common English words that kern to exactly those lengths. Then he Googled those and found your FAQ page.”

Zero Space: Patrick came up with the idea of using the wordplay/definition split to index into the answer of a cryptic clue, which I thought was super creative. While trying to implement his idea, I wanted to motivate it somehow because it’s a big leap by itself, so I came up with the idea of using two-word phrases to add the intermediate “DO IT AGAIN USING CRYPTIC CLUES” message.

The Superbowl: This is one of my personal favorite puzzles of this hunt, mostly because I really like how elegant the final extraction is. When I testsolved it, I tried looking at various first letters for a while before hitting on the aha.

Very Fun Logic Puzzle: This puzzle was written about a week before the hunt because we decided to ditch a puzzle last-minute, so I have to applaud Seth for quickly coming up with a really nice idea. The first versions of this puzzle were very rough around the edges (the “People live next to the people they like. They do not live next to people they hate.” clue wasn’t there which made the like/hate clues confusing, the title was originally “Verily, Thou Dost Protest Too Much” which suggested a Shakespearean connection, there were several extraneous clues, etc.), but we were able to tighten things up after some testsolves. In particular, thanks to Colin for providing some great feedback.

I think the day before the hunt, Seth discovered that the solution wasn’t unique; it turned out that I screwed up transcribing which clues were extraneous in my testsolve. I fixed the clues in question and double-checked the solution again for good measure, until I was certain that it was clean. Overall, I am very happy with how this puzzle turned out, especially given the rushed nature!

How to Best Write an Essay: Most of what I wanted to say about this puzzle is already in the author’s notes, but I’ll add that I hope this puzzle was educational, in addition to being fun. 🙂 I know that people often accuse Strunk & White of being overly prescriptivist, but I do believe that some of their style guidelines like “Put statements in positive form” and “Omit needless words” are useful to keep in mind. Also, thanks to Brian for suggesting some edits to make things more unambiguous.

A Glistening Occasion: If you thought solving this puzzle was a slog, consider the following: I listened to “One Shining Moment” 10 times when testsolving this puzzle. Later, I was assigned to factcheck this, so I had to listen to the song 10 more times, except pausing every few seconds to ensure that the people only appeared once in their corresponding videos. While fact-checking, I actually discovered a repeat, so Josh changed some names around to fix the issue and I listened to it some more times. To be honest, I think the song is not so bad, but maybe it’s just Stockholm syndrome.

Anyways, I’m still not sure how to feel about this puzzle. When testsolving, it certainly took a long time to watch through the videos, but I had fun trying to find the people (I think it’s the same kind of fun as with Where’s Waldo? books). I think this would have been perfectly fine in the Mystery Hunt, but for a smaller event, it might be a little too sloggy to make teams do.

Puncturing Sensation: This is also one of my favorite puzzles, for a similar reason as The Superbowl. I guess I just really like puzzles where extraction is more than indexing/diagonalization/read first letters.

Television: This puzzle opened my eyes to how good Eurovision songs can be. Even though the extraction is kind of standard, I really enjoyed solving this because the research step is a lot of fun.

Retinal Variants: I just want to say that Abby’s art is awesome, and the puzzle is also a lot of fun. It is a little unfortunate that the mostly likely place for a team to get stuck is on the first step: it would have been ideal if BLAST could have been clued in the puzzle somehow as opposed to oblique hinting in the flavortext, but there are already quite a lot of constraints on everything.

X-ray Fish: If you haven’t checked out the solution page for this puzzle yet, you should do so.

Scramble for the Stars: We were sure going in that this puzzle would be extremely hard, and we were a little scared that it might be too hard, so we were pleasantly surprised when 10 teams solved it within 24 hours. This was one of the few puzzles that we didn’t have any 100% clean testsolves on: the best testsolve we had was when a team of Brian, Rahul, and I were able to finish after asking a yes/no question at around 2 hours, namely “Are the names of the constellations relevant?” However, we decided that this had to go in the hunt because of how ambitious it was.

On another note, the first version of this puzzle was constructed by hand by Lewis and had some truly awful fill (I remember that Hydra was TORGHATTEN ASA, a Norwegian shipping company), but even then I was kind of dumbstruck when we got the aha. As mentioned in the author’s notes, I need to give big props to Jon for making this puzzle at least 5x cleaner, though I also have to say that Lewis is a mad genius.

Famous by Association: This puzzle got some polarized feedback, which makes sense because the pairings are not 100% consistent/unambiguous. Personally, I enjoyed testsolving this and laughed at quite a few of the puns, especially Goldfish which I think is brilliant. I also found it mostly unambiguous, though this is a subjective opinion.

During testsolving, one suggestion brought up was to get rid of the decoy companies. At the time, we didn’t want to do that because we thought it would make it too easy/fraudable (if there’s only one company name related to music, then you don’t really need to know what a16z is to make the pairing), but perhaps the puzzle could have been written to avoid this issue. In retrospect, it might have been a good idea to remove the decoy companies even if it made some of the matchings trivial, because it’s already supposed to be an easy puzzle and solvers can enjoy the puns regardless.

Watchers and Fliers: I’ll admit it: I had no clue that these symbols existed before doing this puzzle. This was made even funnier by the fact that when I began working on this puzzle and looked at the flavortext, I joked that it was going to be about washing machines. Little did I know…

The Treasure of Apollo: So as most of you probably know, we kind of screwed this up. In particular, the most egregious part of this puzzle is that we provide enumerations for the song titles, which are already easily looked up, while you need to index into the character names, some of which are ambiguous. Of course, both of these issues could have been fixed by simply providing enumerations for the character names instead! There’s still a little more ambiguity on top of this in some of the matchings, but I think that’s a lot more manageable. Here’s some more backstory on this puzzle (copied from a Facebook comment I made):

Basically, the puzzle went through a bunch of iterations and failed testsolves (at some point, instead of lyrics in a grid, it was “summaries” of lyrics, which was basically unsolvable if you didn’t already know the songs). Many testsolvers had complaints about the ambiguity in matchings, so characters kept getting switched in and out. At some point, we felt bad that we were making the author (Seth) change it so much, and we were also running a bit low on time, so we just finalized it and pushed it out of our minds. In the end, it was one of a few puzzles that didn’t get a clean testsolve, and that really showed: I think if we had gotten one more testsolve, we probably would have caught the issues.

I still love the idea of matching fictional characters to D&D classes, but I am sorry that we didn’t do better on the execution. 😦

A Basic Puzzle: Oh boy. I don’t think any of us were expecting this puzzle to turn out as hard as it did, since it got two clean testsolves, both with reasonable times. In particular, Nathan soloed it in about 2.5 hours, and five of us (Ben, Colin, Jon, Lewis, me) collaboratively solved it in about 1.5 hours. I’ll share our solving path, which I thought was quite reasonable, though others might disagree:

  • First, per the title, we converted all the numbers to base 2, and found one of them had a suspiciously large number of 1s separated by 0s. Counting the number of 1s in each group, we saw that it almost spelled out OCTAL.
  • We then made a table of all numbers converted to all bases 2-9.
  • In the base 3 column, one of the numbers had a suspiciously small number of 2s. There were 25 binary digits between these 2s, and putting them in a 5×5 square spelled out NONARY (after adding some small number).
  • We gradually picked out more suspicious patterns in the table, like one of the base 9 numbers having 8s every third digit, and found more encodings, like semaphore for the base 9 one.
  • The semaphore spelled out HEXATRIGESIMAL, so we also converted all of them to base 36 and found the one that spelled out DECIMAL.
  • Eventually we got enough letters to onelook the answer, though we also ended up figuring out all of the encodings to check the puzzle’s correctness.

I thought that many teams would do something similar, but unfortunately it looks like this didn’t happen, given that it only got 14 solves in the first 24 hours. I do know that from the hint requests we got, quite a few teams converted the first number to base 36 and found DECIMAL almost spelled out, which caused them to think that all of the numbers would directly spell out words when converted to different bases, as opposed to using some encoding. Another possibility is that teams just didn’t want to take the step of converting all of the numbers to various bases, because it is a little intimidating and kind of annoying if you don’t like computer programming. A third possibility is that teams were intimidated by the low solve count and decided to work on other open puzzles instead. I’d be interested if you have any other thoughts on this.

I had a lot of fun testing this puzzle because it’s like 11 mini-puzzles, each with its own satisfying aha moment, and all of the encodings are fair except for the keyboard one, which is admittedly kind of silly (in our testsolving session, Lewis got that one because he remembered the SUMS puzzle from a while back).

Stephen’s Speed Run: I’m a little conflicted on this puzzle. On the one hand, I loved the challenge of optimizing each level’s solution. On the other hand, you can pretty easily wheel of fortune the last 7 characters of the cluephrase, which isn’t a bad thing in itself, but the first level being one of the trickiest levels probably led to lots of frustration. It is also a little unfortunate that all of the levels were taken from the game, but I do think the optimization challenge lends another dimension to most of them.

For me, I think the “ideal” version of this puzzle would be a challenge to optimize several levels, and the game would tell you whether your solutions are optimal (unfortunately, this can’t be done in PuzzleScript without being somewhat transparent). Once you optimize all or most of the levels, the game would just directly give you the answer. I think something like this would have better captured my favorite part of the puzzle, but it would have been a lot more work, and Nathan was already quite busy. In particular, PuzzleScript was used in the first place because Nathan had already written a Stephen’s Sausage Roll engine beforehand, so he repurposed it for this puzzle.

dʒʌmbəl: From the feedback we got, this was one of the most-liked puzzles of the hunt. This lead to a lot of internal discussion on whether we should try to have more puzzles of this sort, where it’s obvious what you have to do and the work/fun is in doing it. I suppose I’ll save this conversation for another post, though.

YOLO Queue: This turned to be one of the lesser-solved puzzles, which is a little unfortunate, because it’s just so funny after you realize you have to take the Q abilities (I laughed so hard upon discovering Day Waste. Imagine if that were an actual League of Legends ability!). In retrospect, it might have been better to make some of the clues more direct: for example, if “Wow, you really dropped the ball on that one!” mentioned American football somewhere so it would directly clue FUMBLE. I think the only “direct” clue is “No, you’re not playing Zerg, and you don’t need another supply unit”, which is unfortunate because you need to have some Starcraft knowledge for that.

Drive: This puzzle was kind of a pain to make, especially with the SOTHPAW disaster mentioned in the author’s notes. I got kind of lucky with construction, in that I didn’t originally plan for all of the license plates to fit exactly in an 8×8 grid, but was able to fiddle with the plates at the end to make everything work out. Also, thanks to Ben for helping with research (we compiled a list of ~80 license plates before trying to fit them into the grid).

Going into the hunt, I wasn’t really sure how this puzzle would turn out. There is an element of “how was this constructed?!” which is always nice, but I was worried that it wouldn’t be that fun because there is quite a lot of data you can look at when it comes to movies, cars, and paths in a grid, and I wasn’t sure whether the license plate aha would make up for all the possible wrong paths that teams can try. From feedback, it seems like people mostly had fun with the data collection and several people really enjoyed the aha moment, though there were also several who thought it was unfair/underclued. Overall, I’m happy that it turned out at least okay.

Unaligned: Constructing this puzzle was also kind of a pain, especially because we wanted to make sure that none of the 3-letter words in the grids overlapped with the 50 3-letter words formed by rotations, and these 50 3-letter words were also fairly constrained themselves. I am glad that we managed to make the grids without using too much crosswordese, and I’m very grateful to Jon and Colin, who helped a lot with the gridwork: this was certainly a team effort.

We intended this to be a long but not difficult puzzle, with several small steps. The feedback we got was fairly polarized though, and I actually think that this might have been the most polarizing puzzle of the hunt along with Scramble for the Stars. On the positive side, this comment from A Dragon, A Duck, A Llama, and A Panda pretty much sums up what I hoped most teams would feel:

“…I love cool methods of extraction. However, answer extraction is evil, and I’m completely against ever being hopelessly stuck on it. Therefore, I liked the way this puzzle worked — there were many, many small steps towards extraction, but each one was not a very large leap, and each one confirmed itself right away, preventing hopeless stuckness.”

On the negative side, there were several teams who said that there were too many steps or that some of the steps were not that motivated. Some teams filled out the grid without realizing that the clues were rotated and only thought of more general permutations, so perhaps it would have helped to clue rotations more strongly (though I would argue that the fact that the clues aren’t in alphabetical order does imply something more specific than permutations). Other teams just thought that the puzzle overstayed its welcome, which is a totally fair criticism. Personally, my favorite puzzles have always been the longer, more “epic” ones, but perhaps these kinds of puzzles are best saved for the Mystery Hunt, especially since teams can get burned out over a week of puzzling.

Thunk!: I really like this puzzle, and pretty much the only unfortunate thing about it is that you can get the answer without figuring out the last 3 boxes, which are the most interesting ones. As mentioned in the author’s notes, this puzzle was made fairly quickly, so the extraction mechanism ended up less than ideal.

To be more specific, what I liked about this puzzle is that the interactivity adds a completely new dimension to the core logic puzzle, especially in the last 3 boxes.  In general, I enjoy interactive puzzles in puzzlehunts, because they’re usually some of the most creative ones, and the space of interactive puzzles feels like largely uncharted territory still.

Conference 1: It turns out that we “secretly” had metapuzzles this hunt, though we suspected that teams might realize something was up with the slightly weird/long answers. We tried to word our FAQ/Rules pages to avoid giving away information (e.g. we didn’t want to specify that some puzzles were worth more points than others, so we tried to be a little vague in the line “Solving a puzzle will earn your team some number of points. The point values will be provided with each puzzle”), although there were a few people who e-mailed us and asked about metapuzzles, to which we tried to give similarly noncommittal answers. We partly made the decision to include “surprise” metapuzzles because it’s something you can only do once before it’s no longer a surprise. Unfortunately, I don’t think we were expecting so many teams, especially less experienced ones, so having a day that you can’t approach unless you’ve already solved enough puzzles was not ideal for these less experienced teams. In the end, though, I hope people enjoyed the novelty.

Conference 1 was the hardest meta in testing and was also the hardest in the hunt. There are many possibilities for the “connect the dots” part of the puzzle, and there were also a few unintentional red herrings like the word PLATO in the middle of the grid. In retrospect, it would have been better to make a grid where each letter appears exactly 5 times (so maybe a 13×10 grid, or a 10×10 grid with 20 unique letters, or even a 6×10 grid with the 12 unique letters among the words PERSON, AMENDMENT, etc.), as that’s both more elegant and also hints at which “dots” to take. I suggested this at some point, but we were already focusing on other parts of the hunt by then, so we all forgot about it until someone mentioned the same idea in our feedback form.

Overall, I still like the “connect the dots” aha, but it certainly could have been motivated better.

Conference 2: Although I don’t think we originally intended it, it was good that the meta with the most answers was also the meta that you could most easily associate answers with (hidden words are a fairly common puzzle trope and can jump out in a list of answers). I have to make the confession that, while we did test the metas individually, we never tested the “splitting up answers into 3 sets” part, as pretty much all of the regular testers were already spoiled on some subset of the metas. So, I was very happy to see several teams finishing all of the metapuzzles within 24 hours.

I don’t have too much to say about this meta, but I will mention that before making it, I didn’t know anything about what plastic or cement were made of. Also, Jon basically came up with all of the mechanics, and I just contributed some ingredient pairs.

Conference 3: For some background, “binary with hands” was one of the first things I wrote in my puzzle ideas document which I started in high school, so I’m glad that I could make this puzzle 5 years later! We used R-Z in the puzzle titles instead of A-I partially because we really liked both “Stephen’s Speed Run” and “Watchers and Fliers” as titles, but we also hope the X and Z jumped out to suggest using titles.

Duck Quonundrum: As mentioned on Dan Katz’s blog, the “transform a word/list of words according to a long list of instructions” has already been done quite a bit, but we went with this for basically the reasons he stated (it’s a very flexible mechanism). I think Rahul did a really good job with the answers he was given, especially with the clever ending and the very appropriate answer.

The Final Bracket: As mentioned in the wrap-up, making this puzzle was somewhat unusual because we already had two of the three metas written and had to finagle some answers around to get the extraction to work. It was very lucky that we didn’t need to change any of the Conference 3 answers, as those were the most constrained.

As a sidenote, we discovered several funny aspects of the bracket, such as VICTORY CELEBRATION almost making it to the end before losing to ME MYSELF AND I, and SECOND PLACE IS THE FIRST LOSER losing in the first round.




Whew, that was a very long post. Congratulations if you managed to make it all the way down here, and if you have any opinions on any of the puzzles, feel free to leave a comment and I’d be happy to discuss anything. I’ll make one more post with more general thoughts about the Galactic Puzzle Hunt, so look out for that in a few weeks.


Written by qzqxq

May 1, 2017 at 6:32 am

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: