Log in

No account? Create an account
15 January 2012 @ 01:23 pm
Quick Mystery Hunt Wrap-up  
The early reviews are in, and I had a tremendously fun time remote solving the Mystery Hunt this year. Even though I missed the first many hours of Hunt as I was at work, and even though I was in bed at a normal time after we finished on Saturday before midnight (the pre-runaround part anyway), I felt I had a full experience from this 108+ puzzle event and I want to thank everyone at Codex for the show they put on. And when they've solved the long-standing problem of posting puzzle and solutions within moments, not months, of the end of the Hunt, Codex has sealed the deal on a PHENOMENAL job!

You'll hear absolutely no complaints from me that the Hunt was "short" even though I've already seen and heard those from a few others. When the puzzles and metapuzzles are smoothly designed and don't grind to a screeching halt, modern teams at modern sizes can get through them very fast. There may be a disconnect between what I'll call the "puzzle solving clock", each puzzler's individual experience of the Hunt focused on the individual puzzles, and the "big picture clock" which few solvers directly observe until it strikes midnight and everyone notices. This "big picture clock" can't have an extended lifetime with open team sizes without either getting to 200+ puzzles (NO!NO!NO!NO!NO!), which is a lot of effort to functionally change the lifetime of no one's "puzzle solving clock", or really messing up in the design/difficulty of a meta puzzle along the way which simply causes a Hunt to not end. You could run the Mystery Hunt indefinitely if you just made one piece a complete bottleneck by being excessively hard/unsolvable or unavailable until much later. Imagine this meta result: "Watson 2.0 will only be satisfied if you provide at least one answer to an open puzzle on this page." That could take awhile, even for a large team. Many past Mystery Hunts, both successful and not, have hit upon this style of hunt-lengthening solution even if unintentionally (and it is the exact structure of the most successful puzzle hunts in terms of participation from the Miami Herald and Washington Post). So until the institutional feel of "coin is found, Hunt is done" is lost, I'm glad to see this compromise where a hunt simply runs until the announced Sunday end time, a trend we (speaking for Evil Midnight) started with the Hell Hunt many years back; the coin being found at 10:27 PM Saturday after the final meta was solved for that team by 6 or 7 PM is just fine with me.

On Remote Solving:
Maybe a part of my enjoyment with this Hunt came because I felt I contributed a lot more to my team than last year which was my first remote experience and where I didn't have the right mind set to overcome that handicap. Some of that improvement came just from more experience solving with Luck, Stuck on Two Open Puzzles and knowing my teammates better, but some came from doing my remote solving experience this year co-located with another teammate in the Bay Area. Just having two people in a room at times to talk out steps when slightly stuck made things much more efficient. Veep should take some small margin of credit for almost all of what I discuss here as he sanity-checked me many times, even when I'll say I was "working alone" on it. I'd like to think I helped in small ways with what he was doing too.

And I've found a good remote solving mode which seems to be to not invest too often in just one puzzle, except ones truly built for me, but instead to help build good solving frameworks in the shared workspaces so that progress then happens more naturally and productively as time gets invested. Then I can crank on data to fill in the page, just as everyone else can, but those working alongside me in those spots can be on the "same page" to crank on data too in the framework that already fits what I think mechanism/extraction will be. In this way, I can work on Two (or more) Open Puzzles, fitting our team name, even though the puzzle sign-out system Luck uses is meant to limit being too diffuse in effort. How to not break the team guidelines? I simply didn't sign out many things. On the few times I did, I certainly was dedicated like the 30 minutes devjoe and I did I Teach You. I don't know if I could be on-site and be as scattered at times without being called out for it, but I'm coming to view my best contributing mode as minimizing "crank out data" time to 20-30% where I've been at 60-70%, and replacing that with more "meta to puzzle" time - not to be confused with metapuzzle time which will always be at least 30% of my thought time if not my work time. Amazingly, this transition almost exactly matches some reassessment of my scientific working time as I'm growing into the requirements of my new job.

I also added a simple but effective move this year: "When I'm not convinced of your data, I'll put a ? by it." This doesn't sound like a big breakthrough solution, but trust me it is. When remote solving for a primarily on-site team, with "data" being the precious work product, you can view what is typed into a shared resource as being in the virtual equivalent of ink, with just other holes to also fill in with ink, when really you need to erase and edit things without necessarily being destructive of work product. Only where I was really out of my knowledge space, on Headstones, did I fail to look for and fix incorrect data this year. In other cases, like the Now I Know my ABCs puzzle where I first built a spreadsheet framework of letters to notes as described above after having gotten that Aha myself, I placed my ? by both early and late IDs that didn't pan out to my personal humming experience. And if my ? got erased by a teammate I'd put it back up after my own second check if I still knew it was wrong. And then it never got erased again until the correct replacement went in. So if teammates want to know who was anonymously adding to their spreadsheets unannounced at times, adding text in highlighted squares, or adding ?'s, that was very often me.

On the puzzles:
While it is hard to catalog all the puzzles I was part of, I can speak to some of the more memorable moments and solves. I really enjoyed the team co-solve of Paper Trail, a beautiful diagramless with nice pay-off. I also really, really enjoyed going through Winning Conditions with veep. While other members of the team had struggled with what to do there, when we got to it we put together some good hypotheses and followed a natural but enjoyable discovery process of learning this particular black box. Not at the level of my favorite "black box/guessing" puzzle, "Thinking of a Number" from MSPH14, but that is one of my all-time favorites and would have won "puzzle of the year" last year if I bothered to write up my awards as I did the last two years, or even write about our simulcast hosting experience of MSPH14. (I don't blog as much about these events as I should which is why this update is today.) Still, "Winning Conditions" involved learning rules of an undefined guessing game in very interesting ways, and will be on the short-list of puzzles I remember from this Hunt for the fun Aha experiences of the solve and plotting out the approach to get to the end even after we felt we had gotten near the answer. 20 Questions looked interesting to me for the same reason, but the puzzle set-up meant that was better for an on-site group to look at so I didn't get a chance to explore it.

I missed the first logic puzzle set, but did solo-solve the Sudoku/Star Battle, Kakuro, and 3D maze puzzles in this Hunt. This strikes me as a higher number of solos than I've ever had before, and it's not necessarily that I sought them out but we always had enough puzzles open compared to our smaller than many team size that I could lead these to conclusion. And unlike usual, I didn't get stuck at the extraction, such as on the Equal Billing puzzle, where 3 grids and 6 letters came together for me cleanly enough to get a solid guess out without yet having the stars on the 4th almost solved sudoku. I'm tempted to say the Sudoku and Kakuro ones destroyed me, as collectively they took 3-4 hours of my solving time at least. In the former, having created Surplus Sudoku as a variant in Mutant Sudoku (and a complement to Wei-Hwa's creation of Deficit Sudoku), and knowing lots of tricks with sudoku in general, the solve just took longer than I wanted. Having 2-4 answers to each (without the star part) may have been a part of the challenge. I also eventually needed to apply what I think was an ungiven assumption: that stars wouldn't end up on given numbers and on the three grids I got to completion (the 4th had the sudoku bits but not the star battle finished) that was a correct thing to do. I wonder if other teams mostly used code on these or piled up the man hours more than we did at 2-3 man hours but all of those mine. The Kakuro also felt like it took a very long time as it had a tiny number of starting points to find before slow progress. Veep got to experience most of my cursing as promising digits almost always ran into a ? clue that halted things, but I found these the hardest of the Hunt puzzles I did this year and normally it is not logic puzzles that make me say that.

On the Meta Puzzles:
And then, where I was happiest to contribute, I pushed on metas way more effectively than last year. It's hard to contribute too much as a remote solver when work and discussion is being lost from the on-siters, but by putting our workspace online as much as possible, and using several teammates but particularly zebraboy as a sounding board, I was a part of lots of solves. After no progress on Critic 1, my initial thoughts and observations got our team unstuck on that one and we worked together from there. I even found myself writing simple perl scripts to split the messages by words where coding for puzzle solving is something I haven't done much or at all before and now, as it is more common in my work, I've added a dangerous weapon to my arsenal. I literally destroyed the Critic 3 meta, getting almost all the separate Ahas myself. That meta still contributes for 5 unsolved critic puzzles where we were stuck and from my solve could move on when I got it with just 3/8 complete. Unfortunately backsolving to the very odd answer phrases was not really possible so my breakthrough didn't lead to shortening the puzzle release clock like other metas.

Many of us independently knew what was involved in the Phantom/Operator meta but I did the code to letter guesswork at 6-7/9 answers to get the answer and the backsolves. One of those "Thomas suddenly types the meta answer in the chat window" moments, as I'm working away on a google spreadsheet without much chatter while others are doing other things. Zebraboy and I talked through the Ogre/La Mancha meta pretty effectively, at 5/8 answers, but I'd like to think with sleep we could have gotten where we needed to go even sooner. Certainly my favorite of the meta answers.

And I had the framework and categories for Watson 2.0 really quickly with help from teammates, and used "presolve back-solving" power which wasn't always that helpful as there wasn't much more to unlock but clearing the puzzle board is always of value. I'm disappointed in myself, at trying to "presolve back-solve" unicode foods, that I didn't jump to Hamburger for raw bar, a natural 9-letter answer for a puzzle screaming for a 9-letter food related answer. I saw HAMBURGER and thought "hmm, that's a good 9 letter answer, unfortunately it is not a vowel." We had ?L?[AFKPUZ]F?U? when I first suggested BLUFF OUT (which is incorrect). Then we actually got that F, to have ?L?FF?U? and we finally submitted that choice and FLUFF CUE and other wrong answers. But by then it was "obvious" the third letter would be a vowel. So I forced that desire to let us try SLICE OF PIZZA (U) and WATERMELON (I), but not the much much better feeling for raw bar answer HAMBURGER (T) in that third spot. So if I had any big failing this year, it was not presolve back-solving that one Watson answer when my Snydey-sense told me that was the most interesting of the unicodes I was looking at by far. That it was on a meta that itself involves a computer that selects signal from noise is the sad irony of the whole thing. It cost us probably an hour or more after that unspoken thought to get the meta, after another puzzle solve, and is the only situation this year (more like 5+ last year) when I had a thought and didn't say it that proved correct. And it's not that all I do is speak thoughts randomly. I know those solvers and can't work with them at all. I always try to be on point, but dialing my filter to be more vocal comes with getting more comfortable with my team. Credit to Dart for a brilliant get of the answer when we were out of pattern matching ideas and starting to think about another route to letters in the meta as "BLUFF OUR" + second part like acrostic of round answers was our new working hypothesis until additional solve data came in.

Altogether, a fun musical themed Hunt and a rather enjoyable 24-hour solving experience for me once I could join our team after work. The metas had nowhere near the brilliant innovations of last year's meta structures, and none shines like Mega Man from that year, although the universal dual use of round answer words was cute. Combining free-standing ("pure") and framed ("shell") metas in the structure as is was done cleanly, and Codex deserves some credit for that. But in comparing to last year's hunt, the individual puzzles shone brighter for me this year particularly because I didn't feel "stuck" on anywhere near as many as last year, and none that as to now revealed to have "very arbitrary" extractions that didn't sit in natural low-energy thought wells. I still detest that 4-bit binary we never got from last year on the rock climbing puzzle I remember being my motivation killer in the wee hours of the overnight stretch, and a couple others like it. I take no frustrations like that from this Hunt, and a lot of that is from the great team-work we had on Luck but a lot also comes from the clean design from Codex. Thanks to all the members of Codex for staging a great production, and not a bomb.

ETA: It seems I've followed another blog flaw of mine of titling my entries with the word "Quick" and then typing for over an hour. I hope it doesn't enter TL;DR territory.
jedusor on January 16th, 2012 03:25 am (UTC)
Watching you and ZB co-solve the metas in chat was pretty awesome. (Please tell me "Snydey-sense" is a term you apply to your instincts on a regular basis and not just a throwaway joke for this writeup.)
motrismotris on January 16th, 2012 03:33 am (UTC)
I've used a small range of Snyderman puns before, always when they've made sense given context as it did this weekend.
lunchboylunchboy on January 16th, 2012 03:49 am (UTC)
It is the case that none of the stars ever ends up on a given, though that isn't a necessary condition for solving. I figured teams would work in parallel on the two halves of each puzzle -- unless you were on the team in which case you'd get stuck with the whole thing -- but I don't actually know how most teams approached it. Many of our testers just wrote their own solvers, which I guess is somewhat appropriate since Wei-Hwa's custom solving app, which he kindly lent me, was essential to creating the thing in the first place.
motrismotris on January 16th, 2012 03:54 am (UTC)
Right. My sense was that while it wasn't a necessary condition on your part, it certainly fit the "elegance" mindset of a hunt constructor. So I went with it and when it proved true (but also VERY helpful - such as how it gives the correct of two choices for a column 1 star in the big pink region of one grid), thought I might have missed something hidden in the flavortext or instructions.

Edited at 2012-01-16 03:56 am (UTC)
MellowMelonMellowMelon [wordpress.com] on January 16th, 2012 04:45 am (UTC)
I really enjoyed this hunt too. Being one of 100+ on Manic Sages I didn't exactly get to solo-solve the logic puzzle sets, but I did usually end up getting over half of the steps with some assistance/checking by others, SnapDragon being the other main progress-maker. That happened on Blackout, QED, Eek!, the Kakuro, and the Sunshine critic meta (Sudoku was solved while I was sleeping). Then there was the Dotted Snake that just randomly popped up in the endgame, which I near-solo'd within about five minutes of it being available. I didn't make very significant contributions on too many others, but that's life on a team of our size, not to mention that my skillset on hunts is pretty narrow.

The most miserable part of the hunt for me was the runaround, partially due to our team size but also because parts of it lacked polish. For instance, we figured out a room to go to and sent people there, but apparently Codex didn't realize and the puzzle wasn't there (not sure of the details). The next 30-60 minutes had all of Manic Sages milling around in the building 7 lobby, totally baffled, before we eventually tried again and found what we were supposed to. Some more awkward "Go back to HQ and let the main runaround team work a bit" and "Wait nevermind we're almost done everyone come with us" and "Oops a bit more to do please huddle over here a bit" switcharoos by our team captains left me without any patience left by the end, though they did achieve their goal of having everyone present when the coin was obtained. I think I'm just gonna skip the runaround unconditionally from now on.
scotthandelman on January 16th, 2012 07:03 am (UTC)
I didn't know you were on Manic Sages, otherwise I would have searched you out at the meetup and high-fived you for awesome puzzles or something.

I wasn't really involved with the runaround because I was infinitely more interested in answering phones and visiting teams, but I think you may be mistaken about the runaround. We made sure that the room to which teams were sent was quite possibly the most boring room in all of MIT. Basically, there was not one hiding place for the coin in that room at all, and there was nothing there for them to find. The realization that needed to occur was that the number gotten from Bitdiddle was not a room number, but a password meant to be entered into his device, which gave the final coin location.
(no subject) - MellowMelon [wordpress.com] on January 16th, 2012 07:47 am (UTC) (Expand)
(no subject) - dumble on January 16th, 2012 10:10 am (UTC) (Expand)
Endgame solution - cananian on January 16th, 2012 06:01 pm (UTC) (Expand)
(no subject) - lunchboy on January 16th, 2012 08:10 pm (UTC) (Expand)
(no subject) - gwillen on January 19th, 2012 04:10 am (UTC) (Expand)
(no subject) - motris on January 16th, 2012 03:33 pm (UTC) (Expand)
(no subject) - (Anonymous) on January 16th, 2012 06:59 pm (UTC) (Expand)
(no subject) - lunchboy on January 16th, 2012 07:01 pm (UTC) (Expand)
(no subject) - gwillen on January 19th, 2012 04:03 am (UTC) (Expand)
(no subject) - dr_whom on January 19th, 2012 08:14 am (UTC) (Expand)
(no subject) - motris on January 16th, 2012 04:31 pm (UTC) (Expand)
(no subject) - affpuzz on January 16th, 2012 07:04 pm (UTC) (Expand)
(no subject) - motris on January 16th, 2012 07:15 pm (UTC) (Expand)
(no subject) - MellowMelon [wordpress.com] on January 16th, 2012 11:12 pm (UTC) (Expand)
(no subject) - motris on January 16th, 2012 11:30 pm (UTC) (Expand)
(no subject) - motris on January 17th, 2012 12:09 am (UTC) (Expand)
(no subject) - gwillen on January 19th, 2012 03:59 am (UTC) (Expand)
scotthandelman on January 16th, 2012 07:04 am (UTC)
Thank you for sharing your thoughts and all of us at Codex are very pleased that we were able to please you. I'm sorry that you missed Blackout. It was an idea that I sat on for months until our head editor said "We might be short...do you think you could whip this one up for me?" So I wrangled Alan Fetters' help by saying "At the very least, Thomas Snyder will talk about the puzzle in his blog." Oh well.

It also makes me happy that the Watson 2.0 meta was impenetrable for you with 5/8...our goal was to not leave a bunch of orphaned last-round puzzles that never get solved, and it's tough to come up with an eight-letter answer for something that would annoy a supercomputer and is not extremely guessable (as Bergman unfortunately was). I think ALTFFOUR fits the bill; it's not an answer that can be grepped, but once you think of it, it clearly has to be right.

Again, thanks for your kind words. We knew that we would never be able to top Metaphysical Plant's structures from last year, but we ran with the advice "Your first hunt should be simple and clean. Use all of your craziest ideas for your second hunt." If nothing else, we ran a damn clean hunt: every puzzle frontsolved, and a single spelling error erratum.
(Anonymous) on January 16th, 2012 02:41 pm (UTC)
I think that the Watson meta being impenetrable with 5/8 depended primarily on which 5. ?L?FFOU? makes it look like too many other promising things. I suspect ?LTF???? might have been guessable all by itself. Stupid hamburger. :-)

Speaking of which, thank you to motris for finding the Unicode block with all the foods in it. I'm a font designer, and I *knew* there was a block with all these weird, useless-looking icons in it, but I just couldn't remember where I'd left it.

-- Dart, Luck, Stuck on Two Open Puzzles
(no subject) - scotthandelman on January 16th, 2012 03:49 pm (UTC) (Expand)
(no subject) - motris on January 16th, 2012 03:53 pm (UTC) (Expand)
Emilydumble on January 16th, 2012 10:11 am (UTC)
Thank you for such a positive review! I'm very glad to hear that you enjoyed the Hunt.

-Emily (Codex admin)
lunchboylunchboy on January 17th, 2012 02:34 am (UTC)
Oh, and thanks for the pointer to Thinking of a Number. Very clever and fun puzzle.
thedan on January 17th, 2012 05:19 am (UTC)
Just to give credit where credit is due, I think the "keep Hunt going" scheme originated with SPIES, not with us. But SPIES didn't publicize it much before the Hunt, whereas we tried to push it harder than they did (and tried to introduce some motivators to keep people going).

Glad you had fun! We had a great time on Too Big To Fail, and you were very worthy adversaries.
motrismotris on January 17th, 2012 05:10 pm (UTC)
SPIES was my first hunt so in my wide-eyed youth state I could have missed how it differed from past ones, so I defer to your better knowledge on who did what when.

I still wonder how much better we would have done with more solvers local this year (we had a good amount of solving power off-site). I remember Evil Midnight's magic recipe was everyone solving in the same room, although on Luck we are getting much better at merging in-person and remote capabilities while still being small.
AJDdr_whom on January 19th, 2012 08:24 am (UTC)
SPIES didn't publicize it much before the Hunt because it didn't really occur to us that it was very likely to happen—we were more concerned about what would happen if the Hunt ran long, since the last three had all run to Sunday midafternoon or later.

I was about to say that I didn't think we'd even planned, before the Hunt, to keep operations running if the coin was found early, but looking through old e-mail I see that we did: our written-out Hunt plans included "There is little we can do about running short.... Instead, we will keep our headquarters open after the coin is found for a longer period than is normal, allowing teams to continue to solve and call in answers." But we didn't announce it ahead of time because we didn't think it was going to be an issue.

onigameonigame on January 19th, 2012 01:31 am (UTC)
cyrebjrcyrebjr on January 20th, 2012 12:30 am (UTC)
Red Herring, what?
(no subject) - lunchboy on January 21st, 2012 03:41 pm (UTC) (Expand)