17 January 2007

I hadn't want to discuss the Round VIII meta at all, I figure the Bombers were punished enough during the wait for someone to solve it and we needn't pile on. But there's a disturbing trend among people whose puzzle writing I love and who are on the writing team that this puzzle "wasn't broken" or "wasn't that bad." Yes it was broken, and yes it was that bad! But this is all aimed at future writers, not meant as just nasty stuff about the Bombers. Furthermore, as I said, I'm happy we didn't win. The only thing I'm bitter about with this puzzle is that we didn't install the Wii 6 hours earlier.

I also want to emphasize that it's hard to not have any broken puzzles. Our hunt had several. Some were my fault too. Obviously one should take particular care with metas, but in general testsolving and having only clean unbroken puzzles is really hard, and can only be avoided by eternal vigilence. It's one thing to say it's inevitable that some puzzle will be a bit broken, and it's another not to think it's that bad. Behind the second door lies the potential for really bad hunts.

First let me say there is a difference between hard, broken, and unsolveable. A puzzle can be hard and not broken, and it can be broken and not unsolveable. My big complaint with the West Coast puzzleing I've done is that people seem to identify a puzzle being hard with taking a puzzle and breaking it a little bit. An unbroken puzzle:

  • Should not require multiple simultaneous guesses with no confirmation

  • Should seem "clearly the right thing to do" at least in retrospect

  • Should not require extra cluing. If your puzzle was broken before you added some subtle hint somewhere, then it's probably still broken. (I think this is taken from Setec's list of advice to us writing the hunt.)

Any puzzle that stumps so many smart people for so long is probably broken, even if it is ultimately solveable. And when a good clean hunt comes down to solving one broken puzzle, it is a tragedy. Being stuck on one puzzle is a miserable experience. It's bad enough when it's your own fault (say our team during Monopoly) or when its a bizarre hard to testsolve issue (like the Orange Star), but when it's just due to negligence it's really unfortunate. I hope that the members of Palindrome who are saying this was just hard and not broken rethink things, because it would be sad to have this happen again.

The main problem with this puzzle was that there were a lot of red herrings, and little to confirm your progress. A few very small tweeks would have improved the puzzle immensely. First off, there was no reason for it to be a 10 x 10 grid. That was not used in the puzzle, and led to many wrong ideas. Patterns like that shouldn't be there for no reason. Secondly, the main flaw in the puzzle was that there was no way to know which letter referred to the Jr. senator and which referred to the Sr. senator. This discourages looking at the senator<->letter pairing, and it doesn't confirm the initial guess that it has to do with the senate. We hit on very close to the right idea at some point, but didn't like it because there were 2^50 possible choices, and who would write a puzzle with 2^50 possible choices? (We did pursue this lead, but were unable to find a chart after a while of looking and so discarded it. That's our fault, but that's what happens when you have 100 ideas and the right one doesn't seem much better than the wrong ones.) Something like just having one of them be "NY" and one be "ny" would have made all the difference. Even after hearing that the puzzle used the senate floor, several of us guessed incorrectly how it worked (we thought the jr./sr. split would be given by taking the first or second occurrence either reading left to right or top to bottom, which is at least only 4 choices instead of 2^50, but is still too many choices). Yes it is possible to easily read out the sentence out of the 2^50 possibilities, but the point is that there's no confirmation until the very last step.

Furthermore, it seems likely that this was testsolved without the actual video. If that's the case, then it is very very bad. This is one of the reason why subtle flavor cluing is bad, because it's hard to know the difference between a transcript and a video. It wasn't always easy to tell in the videos how much was memorized and how much was slightly ad-libbed. The other metas all stood alone without in-video cluing. There's a rumor that this puzzle was actually testsolved with a picture of the senate floor, but I have to assume that it was also testsolved later without that, because a team otherwise so competent can't really have been that dumb. Rumor turns out to be false.

Finally, this puzzle illustrates the main danger with shell metas. They don't become easier as you solve more puzzles. For a pure meta, they're almost always easy once you have all the answers. So you can either solve the meta, or you can solve more puzzles. Here solving more puzzles didn't help.

No comments: