Puzzlehunt
I was sort of tentative in the weeks approaching puzzlehunt because I could see, from our team mailing list, that my contribution to the event had been profoundly less than anyone else on the team. (This point was driven even further upon seeing that Andy and Ben had pulled an all-nighter to fix code the night before the event had even started.) And none of the puzzles had my name on them. Sure, living a couple thousand miles away is a reasonable excuse, but it might also be a reasonable excuse to blow off the event completely.
However, there were some compelling reasons to attend anyway. I've always wanted to see how a puzzle event looks from the "other side", so to speak. And the people on the team are all great people to work with. And there also a lingering fear that failing to attend hunt 14 would make me unwelcome for hunt 15.
Arriving at puzzlehunt had almost the same excitement as usual, except the uncertainty factor was the other teams rather than the puzzles. Spent most of the first hour running laps around the opening ceremonies building and taping up posters directing traffic. The opening rules video was a nice production from a team that volunteered to help out. After the opening, I helped staff the puzzle-distribution desk as teams finished their opening ticket puzzle.
(The entire contents of the hunt was later posted online, which was one of several awesome policy changes of the event. Also makes discussion easier to follow afterwards.)
As we got back to puzzle central, most of us ended up hovering around our respective laptops, answering email. Incoming email was diverted into about fifty different boxes based on a batch of Outlook rules. For most of the weekend, I monitored the boxes for Mike's puzzles, since a lot of them ended up being very prone to questions. Especially "Color Me Crazy", the Kinect puzzle, which had a mandatory scheduling component for each team, plus two leaps of logic that teams frequently confused for a bug in the software.
Answering email was, at times, exciting, nerve-wracking, hilarious, frustrating, sometimes many of the above. Some teams brilliantly illustrated the Wizard's First Rule by sending us the answer, not realizing what they were looking at, and asking us for insight about what to do next. Others would try to confirm the validity of data in a completely unexpected way. For instance, if you're trying to score someone on a scale of 0 to 25 correct answers, what do you tell them when they hand you 40 answers with pieces of the 25 broken up among their results? We attempted to use a very strict vocabulary of responses, i.e. "You have X% correct", in order to be fair to all teams, but the gamut of possible questions pushed beyond our expectations. Many team members reported that they would have written their answer sheets differently, in the context of anticipating data confirmation requests. Another awkward question was of the nature "Statement P can be interpreted as meaning either X or Y, which is it?" in a scenario when X and Y both yield the same ultimate answer to the puzzle. (Of course, if they *don't*, then we have a Big Problem, but our puzzle testing was sufficiently good to avoid such disasters.)
As the hunt wore on, we began to add hinting to our repertoire of responses, which came with its own set of challenges. Most of the time, teams would always get stuck at the same place, which made it easy to generate a few prefabricated hinting emails and reuse them. Other times it was not so straightforward. This is another set of issues that would have been useful to build into the puzzle answer page, in hindsight.
Hinting, in general, has been a very controversial topic. Many of the top teams survive the hunt perfectly well without needing hints. Some interpret this to mean that no one should need them at all. Low-tier teams need hints quite frequently though, and when you spend two days and one night out of your weekend, you do it to have fun, and failing to make progress is generally not fun. Others have suggested that teams should be segregated between casual/taking-hints and competitive/no-hints. However, last puzzlehunt proved that this policy can incite a civil war in mid-tier teams, which struggle between trying to improve their position without hints, and not wanting to throw it all away by accepting a hint and dropping into the lower bracket. Thus teams spent more time arguing about whether they wanted a hint and less time actually having fun with the puzzles.
We attempted to compromise this hunt by developing a new hint system:
- We made public the number of hints each team takes. This number has no effect on score or rank, but is provided for informational purposes.
- A team is “near the front” if less than four hours behind the current third place train. If your train is near the front, we won’t give you any hints. This rule expires once three teams finish the hunt.
- Hints are dynamic - you tell us what you’ve tried so far, and we’ll give you a hint for the next step of the puzzle.
Overall this system was fairly successful. It made little difference to low-tier teams, which were likely to get hints anyway, but it added a useful bargaining chip in the mid-tier civil war: teams could see the number of hints of those teams ranked near them, and choose whether to have a moral victory of ranking lower with fewer hints ("it hurts but we're tough"), versus accepting a hint and potentially moving up to have more fun with more puzzles. The penalty of that first hint was not so harsh as dropping them to a lower division; it was just a harmless number that incremented from 0 to 1.
I had anticipated that hosting the event would allow more sleep than playing the event. This was true, but not in the way that I expected. As the wee hours of the morning approached, the volume of incoming mail diminished and life in HQ became easier. However, at the same time, the top teams were coming very close to finishing the event, and that was exciting to watch. I didn't visit any teams' rooms personally (with the exception of harinezumi and cyfis) but we could live vicariously through other teams by watching their stream of data submissions and infer their train of thought on our super-challenging final puzzle. This was cool enough to push my planned sleep time from 3am to 7am, when the second team finished. (Later on, there were submissions like WEVEBEENWORKINGONTHISFORFIVEHOURSANDITSUCKSBALLSWEHOPEYOUREHAPPY)
Various other lessons learned:
- Giving out the phone number of HQ is something to be done very sparingly. The Chameleons team, which eventually finished sixth, was sent our number early in the hunt and ended up calling us over three times as often as every other team. Our phone in the conference room was on speaker, so it was a major distraction from email responsibilities. Taking it off speaker would have required a relay system, since almost always the answer to the caller's question was known to someone other than the one who first answered the phone.
- Make sure the audio-visual guy teaches you how to work the projector *before* the 900 people arrive.
- Location puzzles are really taxing from a cost/benefit analysis. It's a good thing the Kinect puzzle was so cool, otherwise it really would have colored me crazy.
- Every intermediate data stream should appear in a puzzle solution (I'm looking at you, Travel Itinerary).
- Hinting ideas as mentioned above.
Some random puzzlehunt links, mostly for my own reference when visiting this entry later: 1 | 2 | 3 | 4