2019 Mystery Hunt Roundup
Jan. 27th, 2019 12:31 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
sniffnoySo after five years on Donner Party I defected to Plant this year because David Speyer was trying to get a small Ann Arbor remote solving group going. So unlike usual I didn't go trying to recruit people from the house -- yes, despite living here at Escher I didn't go trying to recruit people from it, even though I did go trying to recruit people from Escher last year when I didn't live here! (Because Nat lived here, recall.) Anyway the remote solving group ended up not coming together very much, so, oh well.
I had been kind of thinking earlier of actually going up to MIT for the hunt for once, but, blech, so soon after getting back from New Jersey? No thanks. But hey, if I'm actually living in New York next year like I'm currently intending, I could quite possibly easy go up there next year! Not sure what team I'd solve for -- if I switched back to Donner Party I'd probably be stuck, as one of the few people actually at MIT on a primarily remote team, doing all the stuff actually requiring walking around Boston...
This year didn't go great IMO but it went better than last year. Plant, of course, has an agreement not to win, since we don't want to write next year, but I am not sure at what stage we would have thrown it had we been close. As it was we did not seem very close. Although they don't seem to be putting up stats this year -- though they put up the solutions real fast, wow, I just haven't had time to write this till now -- so it's hard to tell, but we did not seem close to making it to the runaround when the coin was found.
I will admit I also switched to Plant in the hope it would be something of a bigger team. I remember being on big teams and having to spend so little time identifying things and solving crossword clues because there were just so many people to help with that, so most of my time could be spent looking for insights. But when you don't have a lot of people, it kind of discourages actually trying to finish puzzles because, hey, more things to identify. Anyway, I guess Plant was actually not much of a larger team than Donner Party -- or perhaps, like Donner Party, was just mostly diurnal -- and so I spent a lot of time this year doing the drudge work of the hunt.
I mean to be honest I was also a bit off my game -- I missed trying some obvious things that I really should have. Or to put it a different way, I failed to spam spreadsheet columns like I should have, trying absolutely everything, instead just stopping and staring at a few columns. But at least other people spotted this stuff eventually; it wasn't like last year on Donner Party where it felt like the whole team was in a malaise.
Also interesting to see a different team's spreadsheet practices. :P
Also staying up all night like I did was probably a mistake. Usually I end up staying up all night for Mystery Hunt because there's some puzzle I can't stop working on. In this case, though, I made a deliberate decision to stay up all night because, hey, I'd been waking up later than I'd like, may as well stay up all night and reset, right? But doing this as a deliberate thing, well... probably wasn't great for my brainpower, you know? (I am back on more of a diurnal schedule now as a result. It's... weird.)
Anyway, on to puzzles I want to mention!
A Vexing Puzzle -- Not much to say about this one, just a neat simple puzzle.
Herbert West, Animator -- We got the movies (although when I got there An American Tail was incorrectly filled in as something else, I forget what), we got the extraneous Lovecraftian elements, and then... nothing. Yeah, you just can't really do much with this one unless you've heard of Lovecraft Letter, I guess. Well, OK, maybe good use of Google would've helped, but <shrug>. Someone got it eventually, anyway.
Pass da Pasta -- This one was kind of fun to figure out. I totally did not get what was going on with the secondi. It does seem we messed up one or two fo them, so at the end it didn't quite spell "Una pistola per cento bare", but obviously it was not hard to figure that out -- except someone person initially put "Una pistola per cento croci", which clearly fit less well. Oh well. That was fixed quickly.
Side dishes -- This was a tough one. The end was a little poorly done, I think; we had most of "CONFLICTKIMONOEG", but were missing a few letters in "CONFLICT" (which we easily filled in) and also the last letter of "KIMONOEG". So we tried to fill it in as "KIMONOES" (sic), yielding "WARDROBES", or "KIMONOED", yielding "WARDROBED", before finally just trying "WARDROBE". The problem here is that if you have "CONFLICT", "KIMONO", and "EG", there's nothing to indicate that you should read this as two clues -- "CONFLICT" and "KIMONOEG" -- rather than one clue. We of course realized to read it as two, but were expecting the second to be a single workd, not "KIMONOEG".
You're Gonna Need a Bigger Boat -- Just a neat puzzle.
Your Wish is My Command -- I basically did the code identification on this one. At first I was like, what, old Game Genie codes, how are you ever going to find those. Turns out that the official Game Genie website still has them neatly indexed though, wow. I couldn't identify a few of the games but other people could and from there it was easy to get the codes. After that we were stuck for a long time; other people tried entering the codes we found into the ROM, to no avail. I totally didn't think of decomposing the codes into their component parts! That's pretty clever. Should've thought more about things you can do with Game Genie codes; got a little too stuck on the one approach there...
Fake IDs -- Ugh, missed a real obvious clue here. Identified the photos -- first name, middle initial, last name, DOB, sex (although I missed that Karlie Kloss has a middle name; Harrison Ford, meanwhile, apparently does not) and then got kind of stuck. I noticed that all the starting letters matched the last initial except John Cusack, but failed to take this as the obvious glaring hint it was. Instead I was like "huh I guess I got that one wrong... nope, it really is John Cusack", while missing that this should've been a big glaring hint like I said. Instead the theme led me to think, maybe we should find out how old these people were when these pictures were taken? Yeah, that went nowhere. Other people got this one later though...
Just Desserts -- By the time I got to this one people had already realized these were Android versions and written down relevant stats (such as API levels) but didn't know what to do from there. I was the one who started deciphering them alphabetically. Of course, without knowing the theme, it's pretty ambiguous; for instance I intially deciphered the first one as "STROBOX" (which Google tells me is a thing; some sort of software, so I thought it was on the right track). The third one, I figured out that it ended in "GIANT", but what were the first six letters? Nothing quite seemed to work. Of course, it was "THE IRON GIANT", but we hadn't yet realized that each one had a letter missing or what the theme was. So of course someone else wrote down "The poo giant?" :P Anyway eventually I suggested there might be letters missing and someone else figured out the theme and from there it was pretty easy.
The Obligatory G&S Puzzle -- I spent some time identifying people in this one but totally did not get what was going on, certainly not with the first name / last name switch...
Opposites Attract -- Just a neat puzzle.
Bubbly -- What I'm surprised the solution doesn't mention is that determining the winning starting moves is actually even easier than it says. OOPS -- no it's not. Well, I'm going to include as the following paragraph what I originally wrote here, but, well, it's wrong.
The solution given makes use of the fact that, for nimbers *a and *b, one has *a+*b=*(a⊕b), where ⊕ is Nim-addition, i.e., bitwise xor. But the thing is that more generally, the games given here are constructed from taking the game 1 and combining it with itself via addition and colon; and *a:*b=*(a+b). Now the thing is that, on equivalence classes of games, colon is not well-defined. But it is well-defined with respect to the right operand. And notice the way this is structured -- if write each game down as an expression of +'s and :'s, the right operand of each : is always just *1, which is an actual nimber, so the lack of well-definedness isn't a problem. Right, each game can be written as G(F), where F is the pictured forest, and one can define G(F) recursively on forests by G(F1&union;F2)=G(F1)+G(F2) (where this is a disjoint union), and G(F*)=G(F):1 (where F* is F but with a new root attached, with the roots of F as its children). With this, it shouldn't be too bad to analyze each game by hand, rather than writing a program to do it as they suggest. Or, you could still write a program, but it could, y'know, be a simpler program that just uses the above, rather than going and computing mex all the time.
...except, oops, colon is well-defined in the right operand, but we need well-definedness in the left operand, and we don't have that, so the above paragraph falls apart. OK, I guess maybe you are best off doing it the way they suggest after all.
Thanksgiving-New Year's meta -- Man, I and so many other people on the team thought this might be Dots and Boxes. It was not.
...FISH Puzzles -- Oh boy, so, this one. Obviously we started off solving the sub-puzzles (I went straight for FJF,S; I also solved The Resistance (though I could never get the first one to come out right, but the word was obvious at that point) and helped with Rectangles), taking cues from 2015, but then what? The 2015 meta didn't make sense in this new context. We extracted the phrase "NO SUCH THING AS A" (though we never actually solved Twelve Grids of Pictures, IINM), but what to do with that? "FREE LUNCH" was not the answer, but someone else suggested "NO SUCH THING AS A FISH". That wasn't it either. It wasn't until much later going back that I finally thought, wait, what the hell is "No Such Thing is a Fish", anyway? Turns out it's a podcast! And its first episode is "No Such Thing as a Pilot Fish" -- "pilot fish" being what someone had suggested one of the pictures was. Well, with that in hand, it wasn't hard to identify the rest of the episodes... but then what? The depths given clearly looked like times, but nobody who listened around those times could find anything to work with. Eventually someone got it, but, man. I gather some of the words were more obvious; maybe if I'd listened to more than the first few I could've gotten it. I dunno.
Insider Trading -- This one was brutal. Finding the hidden anagrams was fricking tough, given that they could include word boundaries, especially when you didn't know what the categories were. (Maybe I should've written a program.) E.g. one person found "Happy" as an anagram somewhere and figured there must be a seven dwarves one. Nope; it was purely spurious. Even knowing the categories, yikes. I nearly found MOUNT HOLYOKE; I found the right area... but then I noticed it didn't quite work, because I was looking for MT HOLYOKE. I didn't think to slightly expand and look for MOUNT HOLYOKE. Oops. We actually didn't get the Mambo No. 5 one at all; nobody had any idea about that one. So we had "O?X BEFORE LION"; thanks to whoever realized that was "OSX", because that does not look like a word!
Safety Training -- I got to this one and was like, oh, it's Dumb Ways to Die! Yeah, everyone else had already realized that. :P
State Machine -- Ugh, this one. I made a bunch of dumb mistakes here. Someone else on our team (I forget who) had already figured out the basic idea of how to determine which state corresponded to which symbol, but it fell to me to tediously carry it out. After that, well, I figured the text on top basically said what to do -- take the expressions as expressing an affine transformation, iterate it on 0 until you reach something nonnegative everywhere (and not zero). Although, it wasn't clear to me whether to take the diagonal to implicitly consist of 1s or 0s; I initially thought the former, because why would they have explicitly written "+" on the positive numbers otherwise? So I coded it up and, um, it didn't terminate. It fell to Ari Nieh (I think) to point out that my code was wrong, because I hadn't taken care to make sure that all the variables update simultaneously, so I had basically entirely the wrong transformation implemented. Once that was fixed, we had numbers associated to states, and indexed to get letters... but then what? How do we order the states? In order of a Hamilton path of them, maybe? But they're not even all connected! Thing is, that was kind of the right answer, if we'd just tried anyway. Eventually someone (I think Ari Nieh again) suggested just writing it on a map, which worked. Man, this one was kind of embarrassing.
Loaded -- Oh, man, I wish I could remember who it was on our team that solved this one, because it was damn impressive. I was staring at this one for a while and got nowhere. What the hell did the colors mean? For a bit if I wondered if they were related to Formula D dice but that seemed to be pretty clearly not the case. Someone else noticed the message "LEFTOVER DICE", but what do you do with that? Eventually someone came along and tried adding up the numbers; when I came back later, I saw these sums and noticed they were all multiples of 6. So I added a column for what you got after dividing by 6, which, I quickly realized, could be better phrased as the average. I also noticed that in each row at least one die was equal to the average, but that was as far as I got.
The real breakthrough was some other person -- again, I don't remember who -- suggesting that, hey, these purple dice are always equal to the average; maybe there's a purple die in each row? Because in fact there was always, in each row that didn't already have a die colored purple, a unique die colored purple. And with that idea -- that each row contains one die of each color, they're just not marked already, and they don't need to be the unique thing fulfilling the condition but just need to fulfill it -- well, you could do a lot. That same person, whoever it was, suggested green was the minimum, and I pointed out was that red were the dice that rolled the maximum number on the die. So what about yellow and blue? We never got yellow and blue! This was the amazing part -- whoever it was, I wish I remember, managed to put together the answer just from the three possibilities for the leftover dice in each row. Wow!
Valentine's Day-Arbor Day meta -- the basic idea here was known for a long time but we lacked the answers needed to pull it off. At one point someone tried a blind submit of "TINDER", based on the flavor alone (get it? :P ). That was not it.
Connect Four -- Another one where I wrote a program that didn't help. The mastermind puzzles were tough, so I tried writing a program. By the time I finished, all but one were solved. I tried running it on that one and it didn't terminate. Likely I coded it wrong? :-/ Once "SEE RECENT GMPUZZLES" was extracted, I lost interest in this one.
Something in Common -- Basically my only real contribution to this one was identifying which lines were sung to the tune of "Semper Paratus", and also the Mickey Mouse March. But someone else might have done that too; we were kind of disorganized on this one and so due to bad spreadsheet organization there was a fair amount of duplication of effort. The solution is pretty neat, though.
He's Out!! -- We were stuck on this for a long time; I dunno if we ever solved it. The flavor text on was so misleading! Rings, OK, that's boxing, because it involves Punch-Out!!. Diamonds, OK, that's baseball. But spheres? What the hell is spheres? What, is it fricking Blitzball? I didn't see anything in the puzzle to indicate that or anything else that could be spheres. I guessed maybe it was a sport that involves players being "out", at least, like boxing and baseball? Turns out it was a total red herring. Oops. Anyway, yeah, due to the number of them (13), I initially thought this was supposed to be the Wii version of Punch-Out!!, since that's the only one with exactly 13 opponents (excluding Donkey Kong, of course); but, no, it's the 13 fights in the NES version (excluding Mike Tyson), and some of the opponents are repeated. This definitely caused some confusion where I deleted some correct answers because, hey, we already have a Piston Honda listed! (And similarly with Von Kaiser.) Oops. Anyway, yeah, we noticed the ages were wrong but didn't really know what to do with that. I don't think we made it any further than that.
Moral Ambiguity -- When I saw this puzzle, I immediately thought of two possible interpretations: An easy one and a hard one. The easy one -- which I didn't think was very likely, but had to try -- was that each "0.5" indicated literally half a square. If this were the case it would be easy, since you could just duplicate each row and column, multiply all values by 2, and plug into a nonogram solver. I tried this, for what it was worth, but of course that wasn't it, it was the hard one, where 0.5 indicates gray. Although apparently less hard than I thought -- apparently each number indicated either a run of black or a run of gray, never a mix. Well, IDK, I didn't bother trying this one after the easy way turned out to be wrong.
Cubic -- This was another one where I supplied few ideas but did a fair amount of the drudge work. Someone else (I forget who) had already gotten the essential idea of making a graph of shared roots and splitting it into connected components, as well as noticing what was going on with the negative roots, and had already started identifying the component graphs. (I catalogued the omitted positive roots, FWIW, which was not much.) Since I was like the only other person working on this one, I worked in parallel to identify some of the graphs as well.
It got easier once the other person -- I forget who it was, sorry -- pointed out that since they all seemed to be graphs with well-known names, we could identify possibilities for each based on number of vertices (and edges, being equal to 3/2 the number of vertices), and it was rare for there to be more than one possibility in Wikipedia's list. And it's much easier to check for isomorphism against a given graph, then to stare at a graph you drew and go "Huh, what is this?". (For 12 vertices there were a number of possibilities, but those ones were already done except the Frucht graph, which wasn't too hard because it was planar -- that was before I started using the list and just checking isomorphism, but like I said since it was planar it was easy to draw it and then spot it in the gallery. The only other one that had multiple possibilities, after I started using the list, was the Markstrom graph; but I quickly found a triangle in the given graph, and the other two graphs listed, the McGee graph and the Nauru graph, both had girth greater than 3.)
Really, maybe we didn't even need to waste so much time verifying! Maybe we should've just written down the names based on the list and moved on. In any case, we then got stuck for a while. The other person finally extracted "NAMES FOR THE GRAPHS" from my catalogue of the positive roots, which didn't really help. It wasn't until later that someone else came by and suggested indexing into the names with the number of negative roots. D'oh! I'd even made a column for the number of negative roots, right alongside the column for the names. Like I said... failure to try obvious things. Oh well...
Movie Marathon -- Oy, this one. See, this is the sort of puzzle that made me wish I was on a bigger team. It took so damn long to identify all these. Well, OK -- I didn't do most of the identifying! Not by myself, anyway. So -- basically like one other person showed up to this one, pointed out that, hey, some of these are Detective Conan, and then did basically nothing to identify the actual movies (just noting that none of them seemed to fit the word lengths given). So I decided to identify what I could, which wasn't much -- I got "Your Name.", which, yup, wrong word lengths there, and obviously I could identify many of the series but not particular movies. So you know what I did? I messaged Justine and hastily recruited her for Plant! :D She was able to identify a lot of these, like maybe about half. Oftentimes she didn't get exact names, but, it was good enough to find the names. (Though she misidentified the first one -- "Buddha 2: The Endless Journey" -- as something Three-Kingdoms-related.)
Meanwhile I worked on identifying the particular movies from the various series; by this point I had noticed the pattern, that one word in each title had the wrong length listed, and that it was always shorter than the actual length. (Although for a bit I got the direction backwards here somehow and got confused...) I figured that was the "something's missing" in the flavor text -- that one word in each title had to have some letters dropped. But which? I had no idea. And what about all the rest of the movies?
Thankfully Justine came through with the key insight -- that these were all movies that had been nominated for the Japan Academy Prize for Animation of the Year. With that in hand, I was able to get the rest myself, having a list to work from. (Thank you so much, Justine!) But it still took so damn long! And of course you had to find the particular English title the puzzle-writers were thinking of, which sometimes could be very different than the title listed on Wikipedia (which might list a different English title, or even just the Japanese title). (I think the puzzle-writers were taking their English titles from IMDB?) The hardest ones to nail down were the two Rebuild of Evangelion ones -- I realized in my initial pass through that that's what those were, I didn't need any help for that, but just which ones were they? Usually the word-lengths made it clear which movie in a series we were looking at, but due to the titles in Rebuild of Evangelion all having the same format, and the wrong-length word being positioned at the end in both cases, there was no way to tell other than painstakingly searching YouTube for those particular trailers (and one of them was difficult to find, especially with other movies in the series having trailers that were very similar). (Of course, we probably didn't *need* to identify which was which, but, well, that's the sort of thing you end up doing when you don't have a lot of people to help you; it's easy to spend your time gathering information rather than identifying patterns.)
Anyway, with them all it was obvious that there was one missing from each year (excluding 2018 which wasn't covered), but what to do with that? And what do we do with these wrong word lengths? How do we determine which letters to cut out? I had no idea. I wrote down the words that had the wrong lengths but didn't notice anything -- for a bit I thought they fell into groups like seasons, movie-related words, music-related words, etc, but once I had more that seemed pretty clearly a red herring. I wrote down the length differences, but that didn't help at all.
I absolutely did not think of just taking the wrong numbers and indexing into the answer with them. Now I wouldn't call that an obvious thing to try that I missed, because it requires realizing that there's some non-uniformity going on here; that the incorrect lengths aren't incorrect lengths but rather not lengths at all. But again -- failure to spam columns, you know? I was basically the only one working on this (aside from Justine, of course) so that was where things stopped.
Bloom Filter -- Hey, for once I wrote a program that worked! Well, not so much "wrote" as "copy-and-pasted". The first person to worked on this noted that "BLOOM" passed the filter, but they couldn't find anything else that did. I took a look and realized, hey, this thing is implemented client-side, isn't it? So I extracted the code, ran it on both /usr/dict/words and some Scrabble word lists I had lying around, and got a whole bunch more. I combined the results from each list but I think there were still some missing; still, I think the puzzle was written assuming you might not find all the words. In any case that's basically where I left this one -- someone else went and started organizing the words into categories and later more people joined in but I was working on other things (mostly Movie Marathon :P ). I mean, I realized along with everyone else early on that you would have to find 16 other words and then plug them back into the filter to make a picture (although I didn't know to look at the low bit, that's not obvious), but it's easy to realize that's the last step (aside from the low bit part), much harder to get to that last step. Heh, when we started out categorizing things, we figured that "hash" and "bit" both belonged in the same category of, like, computer stuff; I didn't expect they were actually leftovers for a different part of the puzzle!
A Bunch of Ripoffs -- Man, we got nowhere on this one.
Anyway, that's basically what I had to say about the hunt this year. Maybe there were a few more puzzles I wanted to point out as neat but, eh, whatever. Next year in Boston, maybe!
Next time: I, uh, don't have a next entry planned. I'll think of something?
-Harry
I had been kind of thinking earlier of actually going up to MIT for the hunt for once, but, blech, so soon after getting back from New Jersey? No thanks. But hey, if I'm actually living in New York next year like I'm currently intending, I could quite possibly easy go up there next year! Not sure what team I'd solve for -- if I switched back to Donner Party I'd probably be stuck, as one of the few people actually at MIT on a primarily remote team, doing all the stuff actually requiring walking around Boston...
This year didn't go great IMO but it went better than last year. Plant, of course, has an agreement not to win, since we don't want to write next year, but I am not sure at what stage we would have thrown it had we been close. As it was we did not seem very close. Although they don't seem to be putting up stats this year -- though they put up the solutions real fast, wow, I just haven't had time to write this till now -- so it's hard to tell, but we did not seem close to making it to the runaround when the coin was found.
I will admit I also switched to Plant in the hope it would be something of a bigger team. I remember being on big teams and having to spend so little time identifying things and solving crossword clues because there were just so many people to help with that, so most of my time could be spent looking for insights. But when you don't have a lot of people, it kind of discourages actually trying to finish puzzles because, hey, more things to identify. Anyway, I guess Plant was actually not much of a larger team than Donner Party -- or perhaps, like Donner Party, was just mostly diurnal -- and so I spent a lot of time this year doing the drudge work of the hunt.
I mean to be honest I was also a bit off my game -- I missed trying some obvious things that I really should have. Or to put it a different way, I failed to spam spreadsheet columns like I should have, trying absolutely everything, instead just stopping and staring at a few columns. But at least other people spotted this stuff eventually; it wasn't like last year on Donner Party where it felt like the whole team was in a malaise.
Also interesting to see a different team's spreadsheet practices. :P
Also staying up all night like I did was probably a mistake. Usually I end up staying up all night for Mystery Hunt because there's some puzzle I can't stop working on. In this case, though, I made a deliberate decision to stay up all night because, hey, I'd been waking up later than I'd like, may as well stay up all night and reset, right? But doing this as a deliberate thing, well... probably wasn't great for my brainpower, you know? (I am back on more of a diurnal schedule now as a result. It's... weird.)
Anyway, on to puzzles I want to mention!
A Vexing Puzzle -- Not much to say about this one, just a neat simple puzzle.
Herbert West, Animator -- We got the movies (although when I got there An American Tail was incorrectly filled in as something else, I forget what), we got the extraneous Lovecraftian elements, and then... nothing. Yeah, you just can't really do much with this one unless you've heard of Lovecraft Letter, I guess. Well, OK, maybe good use of Google would've helped, but <shrug>. Someone got it eventually, anyway.
Pass da Pasta -- This one was kind of fun to figure out. I totally did not get what was going on with the secondi. It does seem we messed up one or two fo them, so at the end it didn't quite spell "Una pistola per cento bare", but obviously it was not hard to figure that out -- except someone person initially put "Una pistola per cento croci", which clearly fit less well. Oh well. That was fixed quickly.
Side dishes -- This was a tough one. The end was a little poorly done, I think; we had most of "CONFLICTKIMONOEG", but were missing a few letters in "CONFLICT" (which we easily filled in) and also the last letter of "KIMONOEG". So we tried to fill it in as "KIMONOES" (sic), yielding "WARDROBES", or "KIMONOED", yielding "WARDROBED", before finally just trying "WARDROBE". The problem here is that if you have "CONFLICT", "KIMONO", and "EG", there's nothing to indicate that you should read this as two clues -- "CONFLICT" and "KIMONOEG" -- rather than one clue. We of course realized to read it as two, but were expecting the second to be a single workd, not "KIMONOEG".
You're Gonna Need a Bigger Boat -- Just a neat puzzle.
Your Wish is My Command -- I basically did the code identification on this one. At first I was like, what, old Game Genie codes, how are you ever going to find those. Turns out that the official Game Genie website still has them neatly indexed though, wow. I couldn't identify a few of the games but other people could and from there it was easy to get the codes. After that we were stuck for a long time; other people tried entering the codes we found into the ROM, to no avail. I totally didn't think of decomposing the codes into their component parts! That's pretty clever. Should've thought more about things you can do with Game Genie codes; got a little too stuck on the one approach there...
Fake IDs -- Ugh, missed a real obvious clue here. Identified the photos -- first name, middle initial, last name, DOB, sex (although I missed that Karlie Kloss has a middle name; Harrison Ford, meanwhile, apparently does not) and then got kind of stuck. I noticed that all the starting letters matched the last initial except John Cusack, but failed to take this as the obvious glaring hint it was. Instead I was like "huh I guess I got that one wrong... nope, it really is John Cusack", while missing that this should've been a big glaring hint like I said. Instead the theme led me to think, maybe we should find out how old these people were when these pictures were taken? Yeah, that went nowhere. Other people got this one later though...
Just Desserts -- By the time I got to this one people had already realized these were Android versions and written down relevant stats (such as API levels) but didn't know what to do from there. I was the one who started deciphering them alphabetically. Of course, without knowing the theme, it's pretty ambiguous; for instance I intially deciphered the first one as "STROBOX" (which Google tells me is a thing; some sort of software, so I thought it was on the right track). The third one, I figured out that it ended in "GIANT", but what were the first six letters? Nothing quite seemed to work. Of course, it was "THE IRON GIANT", but we hadn't yet realized that each one had a letter missing or what the theme was. So of course someone else wrote down "The poo giant?" :P Anyway eventually I suggested there might be letters missing and someone else figured out the theme and from there it was pretty easy.
The Obligatory G&S Puzzle -- I spent some time identifying people in this one but totally did not get what was going on, certainly not with the first name / last name switch...
Opposites Attract -- Just a neat puzzle.
Bubbly -- What I'm surprised the solution doesn't mention is that determining the winning starting moves is actually even easier than it says. OOPS -- no it's not. Well, I'm going to include as the following paragraph what I originally wrote here, but, well, it's wrong.
The solution given makes use of the fact that, for nimbers *a and *b, one has *a+*b=*(a⊕b), where ⊕ is Nim-addition, i.e., bitwise xor. But the thing is that more generally, the games given here are constructed from taking the game 1 and combining it with itself via addition and colon; and *a:*b=*(a+b). Now the thing is that, on equivalence classes of games, colon is not well-defined. But it is well-defined with respect to the right operand. And notice the way this is structured -- if write each game down as an expression of +'s and :'s, the right operand of each : is always just *1, which is an actual nimber, so the lack of well-definedness isn't a problem. Right, each game can be written as G(F), where F is the pictured forest, and one can define G(F) recursively on forests by G(F1&union;F2)=G(F1)+G(F2) (where this is a disjoint union), and G(F*)=G(F):1 (where F* is F but with a new root attached, with the roots of F as its children). With this, it shouldn't be too bad to analyze each game by hand, rather than writing a program to do it as they suggest. Or, you could still write a program, but it could, y'know, be a simpler program that just uses the above, rather than going and computing mex all the time.
...except, oops, colon is well-defined in the right operand, but we need well-definedness in the left operand, and we don't have that, so the above paragraph falls apart. OK, I guess maybe you are best off doing it the way they suggest after all.
Thanksgiving-New Year's meta -- Man, I and so many other people on the team thought this might be Dots and Boxes. It was not.
...FISH Puzzles -- Oh boy, so, this one. Obviously we started off solving the sub-puzzles (I went straight for FJF,S; I also solved The Resistance (though I could never get the first one to come out right, but the word was obvious at that point) and helped with Rectangles), taking cues from 2015, but then what? The 2015 meta didn't make sense in this new context. We extracted the phrase "NO SUCH THING AS A" (though we never actually solved Twelve Grids of Pictures, IINM), but what to do with that? "FREE LUNCH" was not the answer, but someone else suggested "NO SUCH THING AS A FISH". That wasn't it either. It wasn't until much later going back that I finally thought, wait, what the hell is "No Such Thing is a Fish", anyway? Turns out it's a podcast! And its first episode is "No Such Thing as a Pilot Fish" -- "pilot fish" being what someone had suggested one of the pictures was. Well, with that in hand, it wasn't hard to identify the rest of the episodes... but then what? The depths given clearly looked like times, but nobody who listened around those times could find anything to work with. Eventually someone got it, but, man. I gather some of the words were more obvious; maybe if I'd listened to more than the first few I could've gotten it. I dunno.
Insider Trading -- This one was brutal. Finding the hidden anagrams was fricking tough, given that they could include word boundaries, especially when you didn't know what the categories were. (Maybe I should've written a program.) E.g. one person found "Happy" as an anagram somewhere and figured there must be a seven dwarves one. Nope; it was purely spurious. Even knowing the categories, yikes. I nearly found MOUNT HOLYOKE; I found the right area... but then I noticed it didn't quite work, because I was looking for MT HOLYOKE. I didn't think to slightly expand and look for MOUNT HOLYOKE. Oops. We actually didn't get the Mambo No. 5 one at all; nobody had any idea about that one. So we had "O?X BEFORE LION"; thanks to whoever realized that was "OSX", because that does not look like a word!
Safety Training -- I got to this one and was like, oh, it's Dumb Ways to Die! Yeah, everyone else had already realized that. :P
State Machine -- Ugh, this one. I made a bunch of dumb mistakes here. Someone else on our team (I forget who) had already figured out the basic idea of how to determine which state corresponded to which symbol, but it fell to me to tediously carry it out. After that, well, I figured the text on top basically said what to do -- take the expressions as expressing an affine transformation, iterate it on 0 until you reach something nonnegative everywhere (and not zero). Although, it wasn't clear to me whether to take the diagonal to implicitly consist of 1s or 0s; I initially thought the former, because why would they have explicitly written "+" on the positive numbers otherwise? So I coded it up and, um, it didn't terminate. It fell to Ari Nieh (I think) to point out that my code was wrong, because I hadn't taken care to make sure that all the variables update simultaneously, so I had basically entirely the wrong transformation implemented. Once that was fixed, we had numbers associated to states, and indexed to get letters... but then what? How do we order the states? In order of a Hamilton path of them, maybe? But they're not even all connected! Thing is, that was kind of the right answer, if we'd just tried anyway. Eventually someone (I think Ari Nieh again) suggested just writing it on a map, which worked. Man, this one was kind of embarrassing.
Loaded -- Oh, man, I wish I could remember who it was on our team that solved this one, because it was damn impressive. I was staring at this one for a while and got nowhere. What the hell did the colors mean? For a bit if I wondered if they were related to Formula D dice but that seemed to be pretty clearly not the case. Someone else noticed the message "LEFTOVER DICE", but what do you do with that? Eventually someone came along and tried adding up the numbers; when I came back later, I saw these sums and noticed they were all multiples of 6. So I added a column for what you got after dividing by 6, which, I quickly realized, could be better phrased as the average. I also noticed that in each row at least one die was equal to the average, but that was as far as I got.
The real breakthrough was some other person -- again, I don't remember who -- suggesting that, hey, these purple dice are always equal to the average; maybe there's a purple die in each row? Because in fact there was always, in each row that didn't already have a die colored purple, a unique die colored purple. And with that idea -- that each row contains one die of each color, they're just not marked already, and they don't need to be the unique thing fulfilling the condition but just need to fulfill it -- well, you could do a lot. That same person, whoever it was, suggested green was the minimum, and I pointed out was that red were the dice that rolled the maximum number on the die. So what about yellow and blue? We never got yellow and blue! This was the amazing part -- whoever it was, I wish I remember, managed to put together the answer just from the three possibilities for the leftover dice in each row. Wow!
Valentine's Day-Arbor Day meta -- the basic idea here was known for a long time but we lacked the answers needed to pull it off. At one point someone tried a blind submit of "TINDER", based on the flavor alone (get it? :P ). That was not it.
Connect Four -- Another one where I wrote a program that didn't help. The mastermind puzzles were tough, so I tried writing a program. By the time I finished, all but one were solved. I tried running it on that one and it didn't terminate. Likely I coded it wrong? :-/ Once "SEE RECENT GMPUZZLES" was extracted, I lost interest in this one.
Something in Common -- Basically my only real contribution to this one was identifying which lines were sung to the tune of "Semper Paratus", and also the Mickey Mouse March. But someone else might have done that too; we were kind of disorganized on this one and so due to bad spreadsheet organization there was a fair amount of duplication of effort. The solution is pretty neat, though.
He's Out!! -- We were stuck on this for a long time; I dunno if we ever solved it. The flavor text on was so misleading! Rings, OK, that's boxing, because it involves Punch-Out!!. Diamonds, OK, that's baseball. But spheres? What the hell is spheres? What, is it fricking Blitzball? I didn't see anything in the puzzle to indicate that or anything else that could be spheres. I guessed maybe it was a sport that involves players being "out", at least, like boxing and baseball? Turns out it was a total red herring. Oops. Anyway, yeah, due to the number of them (13), I initially thought this was supposed to be the Wii version of Punch-Out!!, since that's the only one with exactly 13 opponents (excluding Donkey Kong, of course); but, no, it's the 13 fights in the NES version (excluding Mike Tyson), and some of the opponents are repeated. This definitely caused some confusion where I deleted some correct answers because, hey, we already have a Piston Honda listed! (And similarly with Von Kaiser.) Oops. Anyway, yeah, we noticed the ages were wrong but didn't really know what to do with that. I don't think we made it any further than that.
Moral Ambiguity -- When I saw this puzzle, I immediately thought of two possible interpretations: An easy one and a hard one. The easy one -- which I didn't think was very likely, but had to try -- was that each "0.5" indicated literally half a square. If this were the case it would be easy, since you could just duplicate each row and column, multiply all values by 2, and plug into a nonogram solver. I tried this, for what it was worth, but of course that wasn't it, it was the hard one, where 0.5 indicates gray. Although apparently less hard than I thought -- apparently each number indicated either a run of black or a run of gray, never a mix. Well, IDK, I didn't bother trying this one after the easy way turned out to be wrong.
Cubic -- This was another one where I supplied few ideas but did a fair amount of the drudge work. Someone else (I forget who) had already gotten the essential idea of making a graph of shared roots and splitting it into connected components, as well as noticing what was going on with the negative roots, and had already started identifying the component graphs. (I catalogued the omitted positive roots, FWIW, which was not much.) Since I was like the only other person working on this one, I worked in parallel to identify some of the graphs as well.
It got easier once the other person -- I forget who it was, sorry -- pointed out that since they all seemed to be graphs with well-known names, we could identify possibilities for each based on number of vertices (and edges, being equal to 3/2 the number of vertices), and it was rare for there to be more than one possibility in Wikipedia's list. And it's much easier to check for isomorphism against a given graph, then to stare at a graph you drew and go "Huh, what is this?". (For 12 vertices there were a number of possibilities, but those ones were already done except the Frucht graph, which wasn't too hard because it was planar -- that was before I started using the list and just checking isomorphism, but like I said since it was planar it was easy to draw it and then spot it in the gallery. The only other one that had multiple possibilities, after I started using the list, was the Markstrom graph; but I quickly found a triangle in the given graph, and the other two graphs listed, the McGee graph and the Nauru graph, both had girth greater than 3.)
Really, maybe we didn't even need to waste so much time verifying! Maybe we should've just written down the names based on the list and moved on. In any case, we then got stuck for a while. The other person finally extracted "NAMES FOR THE GRAPHS" from my catalogue of the positive roots, which didn't really help. It wasn't until later that someone else came by and suggested indexing into the names with the number of negative roots. D'oh! I'd even made a column for the number of negative roots, right alongside the column for the names. Like I said... failure to try obvious things. Oh well...
Movie Marathon -- Oy, this one. See, this is the sort of puzzle that made me wish I was on a bigger team. It took so damn long to identify all these. Well, OK -- I didn't do most of the identifying! Not by myself, anyway. So -- basically like one other person showed up to this one, pointed out that, hey, some of these are Detective Conan, and then did basically nothing to identify the actual movies (just noting that none of them seemed to fit the word lengths given). So I decided to identify what I could, which wasn't much -- I got "Your Name.", which, yup, wrong word lengths there, and obviously I could identify many of the series but not particular movies. So you know what I did? I messaged Justine and hastily recruited her for Plant! :D She was able to identify a lot of these, like maybe about half. Oftentimes she didn't get exact names, but, it was good enough to find the names. (Though she misidentified the first one -- "Buddha 2: The Endless Journey" -- as something Three-Kingdoms-related.)
Meanwhile I worked on identifying the particular movies from the various series; by this point I had noticed the pattern, that one word in each title had the wrong length listed, and that it was always shorter than the actual length. (Although for a bit I got the direction backwards here somehow and got confused...) I figured that was the "something's missing" in the flavor text -- that one word in each title had to have some letters dropped. But which? I had no idea. And what about all the rest of the movies?
Thankfully Justine came through with the key insight -- that these were all movies that had been nominated for the Japan Academy Prize for Animation of the Year. With that in hand, I was able to get the rest myself, having a list to work from. (Thank you so much, Justine!) But it still took so damn long! And of course you had to find the particular English title the puzzle-writers were thinking of, which sometimes could be very different than the title listed on Wikipedia (which might list a different English title, or even just the Japanese title). (I think the puzzle-writers were taking their English titles from IMDB?) The hardest ones to nail down were the two Rebuild of Evangelion ones -- I realized in my initial pass through that that's what those were, I didn't need any help for that, but just which ones were they? Usually the word-lengths made it clear which movie in a series we were looking at, but due to the titles in Rebuild of Evangelion all having the same format, and the wrong-length word being positioned at the end in both cases, there was no way to tell other than painstakingly searching YouTube for those particular trailers (and one of them was difficult to find, especially with other movies in the series having trailers that were very similar). (Of course, we probably didn't *need* to identify which was which, but, well, that's the sort of thing you end up doing when you don't have a lot of people to help you; it's easy to spend your time gathering information rather than identifying patterns.)
Anyway, with them all it was obvious that there was one missing from each year (excluding 2018 which wasn't covered), but what to do with that? And what do we do with these wrong word lengths? How do we determine which letters to cut out? I had no idea. I wrote down the words that had the wrong lengths but didn't notice anything -- for a bit I thought they fell into groups like seasons, movie-related words, music-related words, etc, but once I had more that seemed pretty clearly a red herring. I wrote down the length differences, but that didn't help at all.
I absolutely did not think of just taking the wrong numbers and indexing into the answer with them. Now I wouldn't call that an obvious thing to try that I missed, because it requires realizing that there's some non-uniformity going on here; that the incorrect lengths aren't incorrect lengths but rather not lengths at all. But again -- failure to spam columns, you know? I was basically the only one working on this (aside from Justine, of course) so that was where things stopped.
Bloom Filter -- Hey, for once I wrote a program that worked! Well, not so much "wrote" as "copy-and-pasted". The first person to worked on this noted that "BLOOM" passed the filter, but they couldn't find anything else that did. I took a look and realized, hey, this thing is implemented client-side, isn't it? So I extracted the code, ran it on both /usr/dict/words and some Scrabble word lists I had lying around, and got a whole bunch more. I combined the results from each list but I think there were still some missing; still, I think the puzzle was written assuming you might not find all the words. In any case that's basically where I left this one -- someone else went and started organizing the words into categories and later more people joined in but I was working on other things (mostly Movie Marathon :P ). I mean, I realized along with everyone else early on that you would have to find 16 other words and then plug them back into the filter to make a picture (although I didn't know to look at the low bit, that's not obvious), but it's easy to realize that's the last step (aside from the low bit part), much harder to get to that last step. Heh, when we started out categorizing things, we figured that "hash" and "bit" both belonged in the same category of, like, computer stuff; I didn't expect they were actually leftovers for a different part of the puzzle!
A Bunch of Ripoffs -- Man, we got nowhere on this one.
Anyway, that's basically what I had to say about the hunt this year. Maybe there were a few more puzzles I wanted to point out as neat but, eh, whatever. Next year in Boston, maybe!
Next time: I, uh, don't have a next entry planned. I'll think of something?
-Harry
no subject
Date: 2019-01-30 03:46 am (UTC)no subject
Date: 2019-01-30 06:58 am (UTC)