Tuesday, May 29, 2012

Starting Small, the Puzzle Community, and Puzzle Events I've Never Been To

Or: Thoughts Occasioned by Reading a Blog Post
Every year, I predict that the MIT Mystery Hunt will finally collapse under its own weight.  (Every year so far, fortunately, I have been surprised.)  The Hunt started as a two-sheet page of a dozen puzzles for individuals (or maybe teams), as you can see at the Hunt history page.  Since then, it has grown a little bit every year, as people basically refuse to stop coming.  The most recent version had (so far as I can tell) 130+ puzzles, with teams of half-a-hundred or so (again, so far as I can tell; I've never actually gone myself) participating.  Each year, the puzzles get a little bit harder, and often a little bit more self-referential (you often need to know solutions of prior years' puzzles in order to do one of this year's puzzles), and there's always a little bit more people on the teams.

The other big puzzle event every year is the NPL convention.  I'm not a member of the National Puzzler's League, and so have never been to the con, but it sounds like an action-packed (relatively speaking) weekend: there's a puzzle "extravaganza" (a set of related puzzles leading to a pay-off, like a 1/32nd scale version of the Mystery Hunt), some puzzle contests (both solving and creating) and many many many "unofficial" events and just random puzzles people plunk down on a table.

These two, plus the West Coast-counterpart-to-MIT "Game", give you a pretty strong (IMO) community of puzzlers.  Now, I like to think that I like puzzles, and am reasonably good at them, and perhaps am starting to come around to creating them.  I know (at least virtually) quite a few people who are in this community. The question is then: why have I never joined NPL, and why have I never even considered doing Mystery Hunt?

The NPL answer is perhaps less defensible, so we'll start there.  My first contact with the wider puzzle world was showing up to the American Crossword Puzzle Tournament, which would have had to have been 2001 (I knew of it when I was an undergrad, but spring break schedules didn't line up until I got to grad school).  There was always a plug for the NPL, but it definitely fell flat for me at the beginning--being berated for not wanting to do charades ("Why did you come to Stamford if you don't like charades?!?") was easily written off as just a weirdo, but the nom tradition seemed off-putting (I like spy stories as much as anyone else, and I did my playwriting under a stage name, and I have a "handle" for online stuff, but for some reason I couldn't really get behind noms) and the sample from the Enigma was ... dull.  (I'm a wordy-puzzle kind of guy, but I've never really got the hang of flats.)  I've been told that the Enigma has other things than flats in it, and obviously there's more to the NPL than the magazine, but ... registering has just never seemed to happen.  Every year I think I should, and every year I never quite get around to it.

The answer to why I've never gone to the Mystery Hunt is easier: fear and despair.  Even when I first looked at the hunt, the Hunt seemed so massive as to not be worth my time.  (The first I remember reading contemporaneously was the Matrix hunt (2003), which has about 110 puzzles.)  I could join an established team (of mostly strangers) and not assist much, and maybe only see one-fifth of the puzzles.  I could try to somehow form a team out of thin air, and then maybe only see one-twentieth of the puzzles (as we would probably solve approximately one puzzle in the weekend). Or I could just sit at home and wait for the puzzles and answers to be posted online and gaze in awe, which is what I do.  (I usually make a serious attempt at a few puzzles every year, but I have never yet solved a single one.)  I've known people who have just jumped in to a team even at this late date, but that's not me, psychologically.

This brings us back to the linked blog post above, about "Beginner's Events": where are the beginner's events, these days?  What events are out there that are the early MIT hunts?  There seem to be some candidates: the Post/Herald Hunts of Dave Barry are short six-puzzle affairs that fill an afternoon.  (The last puzzle is released at a fixed time, so that more-or-less removes any time pressure on the first five, which is good from a beginner's perspective.  However, that means that a thousand people are trying to solve the last puzzle at the same time, which can perhaps be overwhelming.)  NPLCon sounds (to me, a non-member) as a fine place to do this sort of thing, as a bit of community outreach, but so far as I know there are no "open" (let alone "beginner") events at the cons.  There are definitely some "regular-difficulty" events out there: I cut my teeth on P&A Magazine and the Puzzle Boat, basically spending a year not solving them until I got my skills up; I did some crossword-based extravaganzas from Eric Berlin and Patrick Blindauer as well as some extravaganzas at ACPT, which is fine for me (as a wordy-puzzle type) but not perhaps desired by the physical-event type.  BLG started out regular-difficulty, although the last set was perhaps a bit more than that.  (And there are definitely the "hard" puzzle sets: Mark Halpin's Labor Day puzzles seem to me to be an extra challenge, and I thought the last Intercoastal Altercations was on the hard side as well.)

Those with more team-solving experience than I can jump in on how well mixed-level teams work; my limited experience suggested that having a "novice" on a team of "skilled" doesn't give the novice much chance to "do" so much as "watch".  This might be a way to get them some experience as to how the mind works, but probably not as good as actually struggling with it themselves.

The ultimate question then is: are beginners' events a good thing?  I think the answer is "yes", as I think we need to have a way for people to find their level and to see improvement over time (other than just failing at things for a while).  Expectations do have to be communicated: a super-team coming down for a puzzle event and finding it to be disappointingly easy will not be very happy (they will enjoy themselves, but only for twenty minutes).  (This is the part I didn't necessarily do very well with my Valentine's Day set of puzzles, but fortunately since it was an online thing people could then move on with their day.)  Can such a thing be done?  I don't know, but I think I will make it my next puzzle (writing) project.  I would like to avoid the "word puzzles on location"-type event, but since I am a wordy-puzzle person, it may end up being rather like that anyway.  Where?  When?  I have no idea.  I would definitely want to sit down and dream up a story-line and some puzzles first (just to make sure it's feasible) before doing any of that kind of planning.  But if you have ideas for where or when, then I'm open to hearing them.  (The obvious place would be something like ACPT, since the average crossword solver is not necessarily big on the creative-type puzzles, but we'll have to see.  I have an idea for wordy-type puzzles that I'd like to bring to ACPT, but I suppose I could try this first.)

Tuesday, May 22, 2012

Bracketology (in a very real and very literal sense)

Or: What Happened To Double Elimination?

Earlier today I was playing in an online tournament that was run using the Challonge! website (for the morbidly curious, you can see the bracket there at the link).  It was a double-elimination bracket, and as I had a couple of bye rounds in a row, I was watching the bracket fill in when I realized a couple of weird things.  First, all the first-round losers (including the "first round second round" games such as mine, where both players had a bye during the first round proper and so played our second round game while the others were playing their first round game) were separated -- after the first round games were finished, no one could play a consolation game as none of the bottom games were filled in.  (It's a little hard to see now in the finished bracket, but if you trace down all the first-round losers you will see that none of them play each other.)  The other realization took a little while: the losers dropped down in a nice neat pattern -- but in every other round (rather than immediately after they lost).

Now for our purposes, that's actually a good thing: the consolation bracket was run using 60%-length games, so getting through two of them in the time the championship bracket could get through one was not a completely outrageous idea.  But in a "normal" situation, this would lead to some sitting around in the top bracket.  I don't remember ever coming across this in my more active running-things-on-a-double-elimination-system days, other than a mostly-serious proposal to FIDE to change their single-elim to a double-elimination tournament in this same two-for-one style (literally, as the main matches were two-day affairs, and he proposed putting in two one-day consolation matches for each).

I had myself, once upon a time (in graduate school), made a program that automatically created brackets for any size group (well, up to 64 anyway).  Granted, it was not fancy on the web (it was definitely a command-line program, which cranked out a .tex file that could be converted to PDF), and the version that I found on my computer didn't have the consolation bracket part in it (I think I never quite got the layout right anyway).  But I remember what I did, so I thought I'd compare what I did (with the losers dropping down immediately) compared to the current state-of-the-art.  I've put a spreadsheet up with the two versions (each in their own sheet) on the web via GoogleDocs, so you can follow along.  (Or, if you happen to need a 32-team double-elimination bracket soon, well, now you've got two to choose from.)  Here I am just comparing the effect of the two-for-one round system, ignoring the effect of how I treated byes differently at the beginning (which is why I am using a 32-team bracket: no byes!).

The championship brackets are basically the same for each (not much you can really do with those).  In the Challonge version, the consolation rounds go as: 16 R1 losers; 8 winners + 8 R2 losers; 8 winners; 4 winners + 4 R3 losers; 4 winners; 2 winners + 2 R4 losers; 2 winners; 1 winner + 1 R5 loser.  There are some advantages here, especially for automated bracket building: every round has a 2^n number of teams (meaning no additional working out of seeding--almost), and in each round where teams drop down from above, there are exactly as many winners coming through from the previous consolation round, so each matchup features one loser from above playing one winner from before.  A downside is that the seeding works a little too well: if you follow the seeding exactly, every single drop-down round will feature a rematch between two teams (so you have to do some fiddling: instead of (say) 15v18 and 16v17, you do 15v17 and 16v18).  In my version, the consolation rounds go as: 16 R1 losers; 8 winners + 8 R2 losers; 8 winners + 4 R3 losers; 6 winners + 2 R4 losers; 4 winners; 2 winners + 1 R5 loser.  You do have the same seeding problems in the second consolation round, plus the "4 winners" round towards the end.  However, you finish in one fewer round (that is, if you can handle all 16 first- and second-round games at the same time, you can finish my schedule in 10 rounds, but you need 11 for the Challonge version).

I thought about running some simulations to see what the types of results would be, but I would need some sort of function that tells me how often seed m beats seed n in a game, and I don't have a well-tested one to hand (or, really, any at all).  I may still do this simulation, with the caveat that it will be for entertainment purposes only.  (Well, I'll be entertained.)  But you can still do a little bit of analysis about the results of the two brackets: the Challonge version definitely rewards winning championship bracket games more than the mine (as a result of the extra byes).  For instance, a team who wins two and then loses (ie loses in the "quarterfinals") would be guaranteed a four-way tie for 9th in the Challonge version, versus a six-way tie for 11th in mine; a team who wins three and then loses (ie loses in the "semifinals") would be guaranteed a two-way tie for 5th in the Challonge version versus a four-way tie for 7th in mine; the losing finalist is guaranteed bronze in the Challonge version, where I only guarantee 4th place.  Also, the teams who go farther in the championship bracket will play fewer games (with the extra byes) in the Challonge version: to win, you must go 10-1 if you lose in the first two rounds, 9-1 if you lose in the third round, 8-1 if you lose in the fourth round, or 7-1 if you lose in the fifth round.  (In my version, you play virtually the same number of games no matter when you lose, if you come up through the consolation bracket.)

Which is better?  Well, obviously I prefer mine, as I don't like standing around and I'm not sure I think that winning championship bracket games should be worth double winning a consolation bracket game.  I don't see a good tournament-related (other than convenience of drawing the bracket) reason for skipping the rounds, but if you can come up with one I'm willing to hear it.

Saturday, May 19, 2012

Black Letter Games Artifact 2: The "Live" "Blog"

I had had the idea to keep solving notes as I went before the second artifact arrived; however, in the heat of battle when opening artifact two I forgot about the plan.  So this is still a bit after-the-fact (although not very much after the fact) and still in narrative form.  When we get to the next one, that will be an honest-to-goodness as-written-down-while-solving experience, I promise.

On to the artifact!

The First Pass

After restarting my "assessment clock" with the code on the package, I look over the card.  The interesting features I see:
  • Hey that's a QR code behind the player!
  • Oh it's that horrid handwriting font again.
  • A bit of gibberish at the top of the back of the card.
  • The stats look all wrong.

The Code

I started by dealing with the QR code.  My phone is not QR-aware (yet, I guess), and I didn't feel like finding out how to make it QR-aware, so I looked for some online decoders, used Paint to cut the code out of the corner of the image (thanks BLG for providing a digital image of the artifacts!) and submitted it to several different online decoders as a check.  The first line read "Throw out the man behind the man in the mask" and was followed by three lines of nonsense.  The three lines suggested the haiku was to be found here, but the second line was very short despite needing the most syllables.  My original interpretation was that I needed to remove "umpire", not that I could really find that in the nonsense underneath.  Submitting that as a check revealed that that was a common misperception.  At this point, I put this aside.

Second Base

I decide to focus on the top where the cryptogram is.  The personal info listed is clearly nonsense.  A quick check shows that there are fourteen numbers listed, and the cryptogram is fourteen letters long.  This suggests that the numbers are a one-time-pad type of key, and the main question is whether the numbers are used to encode (i.e. add the numbers to the real answer to get the cryptogram) or to decode (i.e. add the numbers to the cryptogram to get the real answer).  A bit of back-of-the-envelope calculating shows that the first method keeps all the letters in the range A-Z without having to wrap around, so we begin with that method.  This turns the team QRVQTHBBP GSLKQ into MINNESOTA TWINS.  I don't submit that as an answer, though, since that's ... well, that's a team.  A quick glance back at the front of the card confirms that our player is listed as belonging to the Twins.  I decide this means that this is confirmation that I have done the right thing, but I need to also translate something else.  "2nd BASE" is clearly not gibberish.  "Emmett Earnnell" is also not gibberish, but it does have the same number of letters.  Attempting to perform the same mechanism on the name gives "adebeerscarhop".  I'm not sure what that means, so I leave it there for the moment.

First and Third Base

We now have the batting records information to look at.  The years are given in nonconsecutive order, although they do come in pairs.  I originally focus on the phrase at the bottom "Touching Each Base Counts Towards Letterman Awards" and calculate the total bases for each year.  I only get about three lines done, however, before I notice that all the numbers given are between 1 and 26.  I then scrap that completely and change over to converting the numbers to letters.  I get more gibberish.  I decide to sort by year in an attempt to fix that, but I still have gibberish.  I then remember where I started: each of the 2B, 3B, and HR columns have exceptionally small numbers, so I can convert each number (individually) to total bases.  I then get a series of mostly-well-known players: Joe D., J. Cronin, Yogi B., Billy M., Whitey F., J. Bench, Munson, Cal Jr.  Each of these players has had their number retired (which is clued by the blank column "RET #" on the card, which I did not notice at all until I started writing this up--I instead noticed the huge retired-number symbol while looking up the Wikipedia article on Thurman Munson, whom I had erroneously believed was active a lot earlier than the 1970s listed on the card).  I was planning on turning those numbers into letters, in accordance with the rest of the puzzle, although I became concerned when the first few I looked up were 4, 5, and 8 (since I still had them sorted by year), worried that I was getting the proper ordering and still had to determine what the actual clues were.  Fortunately I got a 15 and another 8, so keeping the letters idea and putting the people back in their original order gave me HOPE and HEAD.  Submitting those answers, I got a confirmation that I had the first and third base answers.  I surmised that the second base answer must come from the top of the card, and I noticed that if I put diamond (suggested by debeer) in the middle I get (somebody) Hope, Hope Diamond, and Diamond Head.  I decided it must be Bob Hope, and submitted those answers rather quickly.  (Too quickly, as it turned out, since the system on the other end apparently discarded my Diamond Head answer and I had to resubmit it.)  As I was waiting on those confirmations, I went back to the cryptogram and checked my work, and found my errors in the original attempt, and was now looking at "A DeBeer's carbon" which definitely clued diamond.

When

After another bit of time spent looking at the results of the QR code again with no new insight, I turned my attention to the comic.  The answer to the "riddle", "On that same date, exactly 100 years after his revenge!" definitely looked like a big clue to the when.  I needed to come up with the "he" that was referred to, who had struck out and then had a revenge.  The picture showed a pitcher, a batter, and an umpire.  There appeared to be an arrow drawn from home plate to third base, with "...AC" written underneath it.  Brainstorming with all of that, eventually I came up with Mighty Casey--I knew there was a sequel or two running around, and I thought third base might represent C, so I had "CAC" to sound out and get Casey from as well.
I submitted "Mighty Casey" as a method confirmation submission, despite the lessons from last time--at least I was worried about it at the time, and I rationalized it to myself by convincing myself that "Mighty Casey" was so unambiguous that it couldn't go wrong.  As I was waiting for the response, I realized that if I followed the arrow (so the word would read backwards), I would have YESAC filled in, and I would have written "YES" in the dots, which seems like a very clever little confirmation to me.  I did receive e-mail confirmation back, looked up the publication dates, assembled the answer (the date of the original publication of Casey at the Bat, with the year 100 year's after the publication of Casey's Revenge), and submitted it.

Signature

I had two seeming clues remaining to puzzle out the haiku: throwing out the man behind the man in the mask from the QR code, and the handwritten "I will possess you!" on the front.  I was completely flummoxed by what the "man behind the man in the mask" would refer to, or what could be possessed; so I decided to ignore both clues and look at the actual lines of letters to see what, if any, patterns would jump out at me.  After a day or two, off and on, the best I had come up with was a series of suffixes in the letters: -matic, a couple -one, a -nova, etc.  However, I couldn't make the whole set of letters turn into suffixes, nor decide  how to pick between all the options for the generic suffixes; and a method confirmation submission came back negative, along with a gentle chide that the hint says to "throw things out", not to add them.

So: it was time to focus on removing something.  The only problem is that I had no idea what to remove.  The lines start with an m, an a, and an n, respectively, so I briefly consider removing those letters throughout, or the letters directly "behind" them, which gets nowhere.  I decide I need to remove a man of some kind, so maybe I need to remove a specific man, i.e. a man's name.  The first line starts with matic, so I try to remove a Matt; there's another t on the line, but it's a long way away with a lot of gibberish in between.  Mack doesn't seem to get anywhere, but Michael does.  If I take out Michael, what's left is "at work or" which is a very good start.  Flipping mentally between "famous Michaels" and "what leaves sensible words" leads me to remove Michael Keaton and uncover "at work or at home".

Next step: Why Michael Keaton?  A small amount of brainstorming (while staring at a picture of a baseball player!) eventually leads to "Batman".  Sure enough, the next line contains Adam West hidden inside it, while the last has Val Kilmer.  The middle line of the haiku (which seemed dangerously short) reads "A Y B A B T U" which takes me several attempts to parse as standing for "All Your Base Are Belong To Us".  I almost replace the initials with the phrase, but at the last minute I remember that this is supposed to be a haiku, and the phrase has eight syllables instead of seven.  So the letters it is.

Overall

The QR code will have to play the "spy story" role for this artifact, I suppose.  Some really neat puzzles here; the second base puzzle I liked, with the slight misdirection and/or confirmation.  I suspect in the final analysis that I am actually missing a puzzle, as presumably there is a "Home Plate" puzzle that leads you to Bob.  (I would guess it would have to come from the part of the card labeled "Catcher", somehow.  My best attempt--and best is definitely a relative term here--is to somehow try to relate "Twins" to "Bobbsey Twins" and if we are being possessive, then we have "Bob's".  Alternatively, perhaps "Bob's your uncle".)  Perhaps you are meant to come up with Bob purely as the only famous Hope, but I expect more out of my artifacts than that (even if that's exactly what I did anyway).  I was not fully on the author's wavelength for the signature puzzle, but I had enough to be able to basically brute-force the rest, so it will do.

Lessons

From this puzzle, the lesson I came up with was "Do things first; rationalize them later".  This is most clearly seen in the signature puzzle (where I needed to come up with one of the people first before I could figure out who the people were), but can also be seen with the CASEY/YES clue (I can't imagine anyone being able to figure that out "forward", so to speak; but once you have come up with the name Casey then filling in the blanks gives you the clue).  I am notoriously bad at these sorts of things (I want all my clues to be "forward" clues), so I need to keep that possibility in mind.