The Mystery Hunt is made out of puzzles
Jan. 25th, 2013 04:40 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
dr_whomAnd here's the post where I mention specific puzzles in the Mystery Hunt that I found worthy of comment, or that raised interesting issues I wanted to mention. This post is much more positive overall than my previous one—I single out a few puzzles for criticism, but overall it's the puzzles that favorably impressed me that I have specific comments to make on, and, as noted, there are quite a few of those.
- The Password Reminder meta (solution). A simple meta, befitting what was essentially a warmup round for the Hunt. I mention it specifically for two reasons: one, I did some wheel-of-fortuning I was pretty proud of to extract the answer from partial data, filling out something like __N _OO _OR ATE to WON TOO FOR ATE. And two, we saw WON TOO FOR ATE, assumed that didn't look like an answer (the meta was supposed to be a password reminder after all, not the password itself), and so we just called in "1248". I was actually all for calling in SIXTEEN. Looking at the solution page now, I see that WON TOO FOR ATE was indeed the intended solution, but Manic Sages accepted 1248 as an answer anyway. I appreciate this choice (especially inasmuch as it never would have occurred to us to call in WON TOO FOR ATE per se), and it touches on something that I feel has been talked about a lot over the past years about potentially being more lenient when the final extraction of a puzzle is ambiguous. In 2011, we had a puzzle that extracted to the phrase TOMMIE'S HAMMERIN' BRO, cluing HANK AARON. If someone just called in HANK, we told them they were wrong—a wrong answer's a wrong answer, no matter how close it is—and waited for them to call back a minute later with HANK AARON. But there's certainly a case to be made that, in cases like that, we should have just said 'Okay, yes, you solved the puzzle, but just so you know, the actual answer is HANK AARON'—since the job of figuring out whether that clue was supposed to indicate HANK or HANK AARON isn't really very motivated or interesting. So the choice Manic Sages made with this meta to tell us we were right when we had obviously solved the puzzle but just not picked out exactly which thing they wanted us to call in was an unorthodox choice, but one that's been discussed a lot before, and I think clearly the right one in this case. It was presumably easier in this case because WON TOO FOR ATE didn't feed any sort of meta, so in some sense it didn't matter whether we thought that or 1248 was the answer. But this approach—crediting a solve without having to psychoanalyze the exact intended resolution of the final clue—seems like something teams may want to consider more seriously going forward.
- The Ocean's 11 round structure. Oh man, I am so in love with this round structure. The way it works is, solving each thief puzzle unlocks the corresponding casino puzzles, but only partway—solving Uncharted Territory gives you a quarter of the clues for Caesar's Palace; solving Random Walk gives you another quarter, and so on. This pushes the envelope on blurring the lines between puzzles, metas, and unlocking in a really elegant and innovative way, and makes for a very new and interesting type of solving experience. The only catch is that it requires that all the casino puzzles be solvable with only partial subsets of the clues, and I have no idea how well Manic Sages accomplished that. (I mean, I literally have no idea; I didn't solve any of these puzzles.) But my team did an amazing job on filling in the grid for Caesar's Palace with only half the clues, and that's a puzzle I'd be impressed to see solved even with full data.
- CrossWord Complex (solution). My only comment on this is, I'm glad we didn't start cutting out the grids and trying to assemble them into a cube before we figured out what was going on. Great puzzle.
- Sam's Your Uncle (solution). Hoo boy. This is the puzzle that consumed my entire Saturday, and we never solved it. It's a great example of one of the problems this Hunt suffered from—it's Way Too Big, with Way Too Many Steps. Just solving the 24×24 cryptic took on the order of four hours, and the relatively weak fill, with lots of obscure words, was kind of annoying to work though. (I recognize, though, that weak fill is a genre convention of British cryptics, and therefore was actually a clue to the puzzle's theme.) When we had the main aha, though, I was ready to forgive this puzzle a great deal, including its colossal size and weak fill—I'm a sucker for "the clue has two answers" gimmicks to begin with; and the replacement of the American answers with their British equivalents just blew me away in terms of elegance, justifying the huge grid and weak fill with this presumably super-constrained construction. But then... it didn't stop. There were another 71 clues to analyze, with letter changes that seemed highly unconstrained. There were multiple plausible orders to put the resulting letters in, and without all of them it was very hard to figure out which was right. We found a series of words in the unchanged letters around the edge, but our correct guess of what to do with them seemed ill-defined and we abandoned it, and there was a C that didn't seem to attach to anything. We were distracted by an imaginary WALRUS in the center ring, which was our own fault because we failed to notice a B in the top right corner. We eventually trailed off of this puzzle in frustration, occasionally coming back to it every few hours over the course of the Hunt to try to wrestle a few more "optional" clues to the ground. This was an extraordinarily ambitious puzzle, based upon a brilliant feat of puzzle construction; and if it had been two or three steps shorter, I would have counted it among the best Hunt puzzles I've seen.
- Space Monkey Mafia (solution) My main contribution to this was basically the instantaneous aha: I looked at the puzzle's title (great title, BTW) and "composers" and said, okay, looks like this one's about patter songs. Then I went off to do some crossword or other, but the idea of a patter song puzzle was too attractive and I was drawn back to this one. Someone spotted "A Few of My Favorite Things", and I revised my diagnosis to list songs instead of patter songs, but close enough, and off we went! I'd never heard Weird Al's "Hardware Store" song before; but you'd better believe that one's getting added to my patter song playlist. The answer extraction here by indexing into "The Elements" was actually one I've used before (when I did it it was basically just a list of indices and heavy flavortext cluing of the song, give me a break, it was the first puzzle I ever wrote), though it was Seth who suggested trying it for this puzzle (after I said '...or maybe we have to use "The Elements" for extraction somehow'). Anyway, then we extracted SING ABOUT COINS and I got to go perform "Both Sides of the Coin", which, being the only patter song about coins that I know, I thought was appropriate.
- Grandson of the Realm of Unspeakable Chaos (solution). Exactly the puzzle I've always been waiting for! Actually I thought about writing a language-change puzzle like this for 2011, but couldn't think of a way to make it fair and solvable; the idea of doing it in Chaotic didn't occur to me (and I would have been a bit reluctant to step on
![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
ucaoimhu's toes, anyhow). Of the three Unspeakable Chaos translation puzzles, in 2001, 2008, and now 2013, I worked on and basically fully translated all three, but this is the first one I actually solved. This puzzle sat right on the sweet spot between solo solve and teamwork solve for me: I worked it from beginning to end and was the one who understood how the puzzle worked and extracted the final answer, but I couldn't have done it without suggestions and contributions at just the right time from bluefaith, noahspuzzlelj, Andrew, and others. A beautiful hard-core philology puzzle; many thanks to the author—though I wonder a bit how many teams that didn't have on them someone with a PhD in language change could have solved this one. I could quibble on a few points—e.g., the merger of all places of articulation to labials before a, coronals before o, and velars before u is just about the least phonologically plausible sound change I've ever heard of—but I recognize that was necessary for the answer extraction and the puzzle played totally fair about it, so why bother complaining? Anyway, I found the actual storyline of the puzzle text and its tongue-in-cheek self-reference quite amusing also.
Actually I think the reason I ended up solving it was the result of a fluke coincidence. I never spotted the "fricatives become stops after high vowels" change; I did see s becoming tʰ in a few places but didn't think too hard about it because 'x-chromosome', the word I wanted to translate, didn't have s in it; moreover I had somehow assumed that final-vowel deletion took place before d-deletion, though I bet the data contains evidence that that's not the case. As a result of these two misunderstandings, I was stuck with lʊɣaʊɣaʊ as my translation of 'x-chromosome' for quite a while (which is why I called in LUAU). Also I didn't understand why tʰrʊtʊp was being used in the text as a translation of 'astronaut', since as I understood the sound changes, 'astronaut' should have been tʰrʊzʊzʊ. But at some point Andrew or Noah said, well, what kind of sound change would be necessary to turn the word you want to mean 'astronaut' into the word that we have for 'astronaut'? And I was like, well, the following unlikely-seeming set of changes: to turn tʰrʊzʊzʊ into tʰrʊtʊp, we'd want to have a rule saying that when you have a pair of identical syllables (zʊzʊ), you collapse them into a single syllable, convert a voiced fricative into a voiceless stop, and add a voiceless stop one place of articulation forward to the end of the syllable, and that's utter nonsense as a sound change. But then I stopped and thought about it for a second, and realized that applying that same nonsense sound change to lʊɣaʊɣaʊ would produce exactly lʊkaʊt—i.e., LOOKOUT—and we called it in. But it turns out the fact that that cockamamie reasoning arrived at the correct answer was a total fluke! The repeated syllables were irrelevant, and the "next place of articulation forward" pattern was a coincidence. Well you know what—I'll take it. - Mergers (solution). This is an example of a puzzle that really needed a better answer, or better editing, or both. Looking over the solution, it seems to me that every pair of clues has a clear, well-defined, and gettable answer... except for the final pair, where the extraction of MAN from AGE and CHESTER seems totally out of the blue and out of keeping with the rest of the puzzle. A simple, fun, and elegant puzzle, but spoiled by a nigh-unmotivated answer extraction; my team worked on this one for a while, considered the correct tree, and then rejected it to try new configurations because AGE + CHESTER didn't clue anything. This violates a cardinal rule of puzzle design—when you have the right answer, you should be able to tell.
- Dear Abby (solution). A Shakespeare puzzle, and I never worked on it or even saw it! But Lanthe solved it and posted on our team chat ACT OUT MACBETH, and I got all excited about performing Macbeth act II scene ii, which I've already memorized, and is about sleeping no more, so it'd be a great scene to perform at 5:00am. But we called that in and were told it was wrong, so we went back to look at our spreadsheet for the puzzle again and noticed that it actually said ACT OUT MACBETH A I SC III, so we did that one instead—everyone who was still awake on our team except chmrr, that is (all four of us), with Barry the Elephant playing Witch 1. ("Which one?")
- Git Hub (solution). This puzzle came up just when we needed it: a clean, simple puzzle with a straighforward but not immediately obvious extraction, for when it was just me and chmrr and Ben on the overnight shift, waiting for the rest of our team to wake up. I don't know the first thing about what a Git repository is, but fortunately the puzzle was actually about the T; and I figured out the answer extraction, which I'm kind of pleased about.
- ...fixes the wavy zigzag jumble (solution). As it turns out, this was a fairly straightforward puzzle with a cute extraction, and ordinarily I love pangram puzzles. Why didn't we solve it, then? Well, first, because it started with a very large crossword grid with weak/obscure fill (making for the second puzzle appearance, and third spelling, of the Asian antelope NYLGHAU) and Monday-level cluing, so before Seth and I even finished filling in the grid we were already kind of bored by it. And second, because we made a mistake solving one of the cryptic clues at the bottom, leaving us at a roadblock without any idea of what to do next with our huge grid; the way the Hunt had been going so far, we were more ready to believe that whatever the next step was was poorly clued and unreasonable than that we had made a silly mistake. And so as a result of both we abandoned what we thought was a dull puzzle that we were stuck on, and that cost us a fun answer extraction. There's a moral of some kind in that, I expect.
Other puzzles that I worked on and enjoyed, but don't have any particular comment on, include the following: The Silver Screen (solution); A Set of Words (solution); Famous Last Letters (solution); The Alphabet Book (solution).
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2013-01-25 10:18 pm (UTC)Casaer's Palace was really tough unless you have all 4 sets of clues. We got most of the grid with 2, but even with 3 the cluephrase was really hard to get because it was in Latin. They tried to get around this by having the cluephrase repeated twice, which I think was a great idea, but it didn't work for us (in part due to bad luck in terms of not having found many overlapping strings, it may have been easier with other sets of three than the one we had). On Excelsior I didn't like the puzzle much, and we got stuck on the third aha, but the fourth answer didn't help us and it was in the "all but extraction" stage when we only had three answers. I can't speak to whether the other ones worked with three out of four.
no subject
Date: 2013-01-26 12:17 am (UTC)no subject
Date: 2013-01-25 11:45 pm (UTC)The team we had on it included the people responsible for Sounds Good To Me, so they were not at all clueless. But I think they got the language changes slightly wrong in some way we don't understand. Or something. It's so far out of my sphere that I am only relaying my guess as to what happened...
no subject
Date: 2013-01-25 11:54 pm (UTC)no subject
Date: 2013-01-25 11:57 pm (UTC)Did you feel like the total workload for the puzzle was reasonable? All I know about it was that our team of 3-4 people were working on it literally from Saturday night through all of Sunday, and I remember thinking "It must be an awesome puzzle for them to have stuck with it. It's probably also broken in some way for them to be stuck on it. Poor saps."
no subject
Date: 2013-01-26 02:32 am (UTC)I don't know if I can judge whether the total workload was reasonable. I think so, but then (a) I didn't find all of the sound changes and got the answer by a lucky coincidence, and (b) have a PhD in language change. But it seemed like a reasonable-but-hard puzzle to me in terms of workload.
no subject
Date: 2013-01-26 12:22 am (UTC)The author of Grandson also writes problems for the Linguistics Olympiads. I think our testsolvers solved it with a lot of hard work and persistence.
no subject
Date: 2013-01-26 01:29 am (UTC)no subject
Date: 2013-01-26 02:33 am (UTC)no subject
Date: 2013-01-26 12:53 am (UTC)no subject
Date: 2013-01-26 02:34 am (UTC)no subject
Date: 2013-01-26 02:39 am (UTC)Glad you enjoyed the puzzle! It's nice to get feedback, especially positive feedback.
Grandson did testsolve successfully (and without the squinting-at-the-answer leap) with testsolvers who knew about a 24.900's worth of linguistics. (Took them about 24 solver-hours, though.)
You're right about the "least phonologically plausible sound change I've ever heard of": it was indeed chosen because it was necessary for the answer extraction and made a reasonable level of puzzle. It actually replaced the rule "consonants become velar if the following two vowels are identical", so it's no longer the least phonologically plausible sound change you've ever heard of. :)
Turning tʰrʊzʊzʊ (well, tʰrʊwʊwʊdo) into tʰrʊtʊp is one bit of evidence that d syllable deletion happens before final vowel deletion (and that both happen before the place of articulation change). As a bonus, that includes defrication of one or two phonemes (depending on your rule ordering) other than s.
-Adam Hesterberg
no subject
Date: 2013-01-26 02:47 am (UTC)no subject
Date: 2013-01-26 03:05 am (UTC)For final vowel deletion before the place changes, there's kaotika > pʰoʊtʰykʰ, not pʰoʊtʰypʰ; tumuparatu > kʰaʊŋaʊpʰorotʰ, not kʰaʊŋaʊpʰorokʰ; prasu > pʰros, not pʰroh; wymu > weym, not weyŋ; and tylipo > tʰeylypʰ, not tʰeylytʰ.
The others can go in lots of orders that always give equivalent outputs.
no subject
Date: 2013-01-28 05:42 pm (UTC)no subject
Date: 2013-01-26 02:19 am (UTC)no subject
Date: 2013-01-26 02:54 am (UTC)no subject
Date: 2013-01-26 02:24 am (UTC)The 1248/WON TOO FOR ATE thing, I totally agree with. Elsewhere (the motris post?) people were suggesting confirmation of partial clue phrases, and I'm not sure I think that's a good idea. But in the case of "You have the answer, but you're not getting our phrasing" seems to me to be exactly when you say "yeah, close enough, here's what we had in mind". Doctor Six from Zyzzlvaria's Reverse Dimension will always epitomize that for me: the final clue was ALFREDO ITALIAN FOOD, and we called in three wrong (but Merriam-Webster-sanctioned) spellings of "fettuccine" before hitting on the one they wanted. That just seems unnecessary; when we called in "fettucini", they should really have just said "Well, OK, yes, and the spelling you'll want for the meta is the following."
no subject
Date: 2013-01-26 02:41 am (UTC)no subject
Date: 2013-01-26 03:11 am (UTC)no subject
Date: 2013-01-26 07:10 am (UTC)First was Sao Paulo. When we called in MAN-O-WAR as our answer, Acme told us "You're the second team to call that in. Can I ask how you got it?" We explained the whole process, that we fiddled with unlikely homonyms like BELLE for BELL and so on, and finally found a way that the words fit to produce an answer. They appreciated our story, then told us that our work was as good as the "real" answer, WALKMAN. Then they called the first MAN-O-WAR team and told them WALKMAN, too. [Side note: The answer key added in 2010 has a completely different answer to the same puzzle. Ha ha. I remember we had STRAY in our solution, as does the 2010 solution, but it was not part of Acme's plan. There's probably something else fishy with 2010, too. An exercise for the reader.]
Second was Capetown. We found the CHEESE early and called it in, but it was incorrect. We found the JAM next and asked "What goes with CHEESE and JAM? Why, a LOG goes right in between, and it's thematic!" Acme couldn't believe we missed the right answer, so we talked with them about where LOG came from, and they told us that we overshot.
Some of this was possible because the Hunt was small enough that all of Acme's phone core could understand the puzzles (or find someone who did) and talk to us about it. Some of this was possible because, well, there were a lot of errata that weekend and Acme was open-minded about things. Whatever it was, I miss it.
no subject
Date: 2013-01-26 05:08 pm (UTC)David S, who was on Setec that year, but is on Plant now.
no subject
Date: 2013-01-26 09:18 pm (UTC)Of course, these aren't quite the same thing--these are solving gone awry, as opposed to (dis)confirmation of solves along the expected path. Regardless, there's a lot to be said for that era, and missing it is reasonable (though you can likely find it in other arenas).
no subject
Date: 2013-01-26 07:26 am (UTC)* Have someone who knows each puzzle on hand looking at the incoming calls. This has obvious logistical and possibly fairness problems that don't need to be elaborated.
* Scrape your database ahead of time for puzzles with slightly ambiguous extractions and populate your system with alternates. This is closer to workable. Still, it has potential traps. Teams will get used to your system doing this, and so when they stumble on the one puzzle you missed they will assume they're on the totally wrong track rather than just calling in the alternate, as they might have normally. A similar thing happens if you don't manage to guess all of the legitimate alternates that are available for a puzzle, and then it's unfortunate for the team that hit the unlucky alternate. And of course, if you try to correct for these effects on the fly, you're back to the logistical and fairness problems of the previous plan.
I'm not saying it can't be done-- I remember the Google hunt a while back had an elaborate, almost AI-like matching system that seemed to work reasonably well-- but it's definitely not trivial. It seems like adhering to good cluephrase practice is the more rigorous and reliable solution, where possible.
no subject
Date: 2013-01-26 09:19 pm (UTC)no subject
Date: 2013-01-26 07:30 am (UTC)<lj-spoiler "Click for SPOILERY solving details.">
By the time I had the text mostly translated, I had the dV-drop, the final-vowel-drop, the vowel shifts, and the devoicing/aspiration rule, and was fairly confident that /l/ and /r/ remained unchanged. I knew that fricatives were changing to stops and places of articulation were changing under certain conditions, but didn't know what the conditions were. Plus, since I always mentally pronounce Chaotic w as /w/ rather than /v/, it never occurred to me to lump w with the fricatives, and so all I knew about it was that (based on forms like zʊpʰoɣaʊnʊlnhaʊl) it sometimes changed to something else. So I had lʊ{?}aʊ{?}, with the {?}s unknown, and was able to leap to the answer from there.
</lj-spoiler>
no subject
Date: 2013-01-26 01:18 pm (UTC)Thanks again for your feedback - I definitely feel like I will be able to put my ideas in much more palatable puzzles in the future.