Setec Astronomy on Writing Hunts

A bit on the management of the hunt - Ray

Mon, 24 Jan 2005 22:09:25 GMT

Each time we've run the hunt, we've had a similar power structure to manage it:

Dictator - Chris. Any decision that needs making, he makes or delegates. His word is final. Try not to argue.

Minister of Taste - Brian Tivol, for the first two hunts. Responsible for knowing how the whole hunt works, including every puzzle, and paying attention to details like making sure lame puzzles or two puzzles that are too similar don't make it into the hunt Tivol moved away, and Chris had to do this for the 2005 hunt.

Asshole - Ray. Holder of the "get out of tact free" card, he gets to tell people the decisions the Dictator has made, or that their puzzle needs redoing, or that we've taken away their word.

I (the Asshole) also complained a lot about how far behind schedule we were, and threatened to take away words from people if they didn't get to work. I also handled much of the data, handling the finalization of puzzles, putting them into the web-release system, etc. I'm pretty sure that sort of management could have been separated from the Assholing, but it provided a handy club to wield, since people couldn't get a puzzle into the hunt without giving it to me, and I didn't have to accept any puzzles that Chris didn't okay.

(I got a lot of help dealing with data, from Tivol in 2000, Tivol and Albert Lin in 2002, and Tanis in 2005, for which I am very grateful.)

More Setec

Mon, 24 Jan 2005 21:58:11 GMT

By the way, if any other members of Setec want to contribute, just ask me for the password. - Ray

news articles about the hunt

Mon, 24 Jan 2005 21:40:56 GMT

Boston Phoenix, before the hunt (Jan 14, 2005).

Boston Globe, by the reporter that played on Illegal, Immoral, and Fattening (I think).

Comment on Square Mess - Ray

Mon, 24 Jan 2005 21:16:50 GMT

Apparently, a lot of teams spent a lot of time on Square Mess. And some of them didn't like it. (This post complains about it having a non-dictionary word, for instance.)

I think this is one of the blind spots on Setec in our testsolving. We don't have a large number of programmers, or programmers that are insane hackers, but we've had this problem before. In the 2000 hunt, every testsolver (there were three separate ones, I think), wrote a program to solve Lions and Tigers and Bears. But (almost?) every team did it by hand.

For what it's worth, I was the first testsolver for Square Mess. I spent between two and three hours coding (it's hard to say, because the first two versions were harder. They didn't use spaces, for instance). But we did have another solver go through it at the end, and he didn't have any trouble. Most dictionaries I checked had all the words (in particular, almost all seemed to have "Warhol", except for /usr/dict/words on athena). I used a large dictionary for testing, but also checked some other sources (like warmerican on Debian).

Perhaps another rule about testsolving is in order: be wary when testsolving puzzles that your team is good at.

Talking back - Ray

Mon, 24 Jan 2005 20:36:05 GMT

First things first: I hate writing hunts. Actually, that might be a bit stronger than I mean it to be. Writing a good hunt is a lot of work. It may not be too bad for individual members of the constructing team, who only need to be concerned with writing and testing a few puzzles, but for the people in charge, it's at least a month of solid work, maybe two. Even spread out over a full year, it's hard to find that kind of time. Chris and I started meeting every Monday and Friday night starting in November, and continued up until the hunt (and we met all day every day in January).

So, it's a lot of work, and a lot of headaches. The payoff is a good hunt and fun for hundreds of people. Overall, I think it's a good thing, but it's time for someone else to give up the time to create the fun.

I created this account so I and other members of Setec could comment about the hunt. There's been a fair amount of discussion about it on livejournal, and some misconceptions about how we construct a hunt that we'd like to clear up. In particular, we'd like to keep future constructors from going down the wrong path. I don't think we've hit on the only way to write a good hunt, but we do have a good way, which has worked three times out of three, so we can claim to know what we're talking about.

I, personally, don't know a lot about writing good puzzles (and didn't like a lot of puzzles in the 2005 hunt that turned out to be big favorites). That's much more Chris's department. But I do know some of the rules for writing a hunt itself:

Testsolve everything, and do it in a way as similar to the hunt as possible (i.e., no communication between constructor and solver).
Hard puzzles should be testsolved twice.
Do not make puzzles harder after testsolving. In fact, don't touch them if you can help it after that point.
Someone has to be responsible for the hunt. For us, the buck stopped at Chris.
Metapuzzles should be solvable with around 80% of the answers.
If the puzzle requires cluing in flavortext, don't make that text obscure. In fact, avoid flavortext if you can.
No more than one "aha" should be necessary per puzzle.
Unordered anagrams are the spawn of the devil. Always include an ordering mechanism. Don't make it too hard to get.
Build the hunt top down.

There are other rules, and rules about individual puzzles and puzzle types. But I think I hit most of the important ones.

More later, from me and from Chris if I can convince him to contribute.