Science Non-Strosity: Discussion (mid-Atlantic: STAY AWAY)

[cvdwightw]

If there is any chance that you will want to play this set at Swarthmore during the ACF Regionals/GARBAGE weekend, please turn away from the rest of this thread. I don't know what's going on with that, but Jeff Amoros has the packets and will presumably be in charge of either running the tournament or finding someone else to run it.

Anyway, thanks to my co-authors (in alphabetical order: Ray Anderson, Bruce Arthur, Matthew Feldman, Darwin Fu, Brandon Hensley, Colin McNamara, Evan Silberman, and Andrew Ullsperger) for helping to complete the set, to Ray Anderson, Bruce Arthur, Auroni Gupta, and Marcus Luna for helping to staff the event, and to Brian Lindquist and Arnav Moudgil for providing rooms and generally allowing this event to happen.

As I posted in the other thread, what I believe to be the correct stats are posted, and you can e-mail me at dpwynne at gmail dot com to obtain a copy of the packets (please specify Word 2003 or 2007 version).

More substantive comments in the next post.

[cvdwightw]

As promised, more substantive comments.

1. I know the Earth Science subdistribution was a little wonky. There was a distinct effort to stay away from "Rock Bowl" and "Epoch Bowl," and we probably swung the pendulum too hard in the other direction with lots of oceanography/atmospheric science/"thinly disguised geography" as opposed to geology. If there seems to be some sort of massive outcry for more geology in the Earth Science distribution, and I'm masochistic enough to head up Science Non-Strosity II, I'll try to do better.

2. I know the Einstein question got really easy really fast. I think that was the only one that had that problem, but there were a few physics questions I was writing at 2 AM on Thursday night/Friday morning that might have had a few problems. If there was anything that was particularly good or particularly bad, feel free to praise/pillory in this thread or by private e-mail.

3. To the best of my knowledge, I did not introduce any factual inaccuracies into the questions, though there may have been some awkward parsing that led to some unfortunate wrong answers, or some vague clues that made it difficult to tell exactly what the question was talking about.

4. I know powers were quite difficult to come by. I erred on probably the wrong side of the "too stingy/too generous" line, as people were clearly buzzing with good knowledge that should have been worth 15 points but was only worth 10. If people really like scoring those extra five points, and I'm masochistic enough to head up Science Non-Strosity II, I'll try to extend the power mark.

5. Despite my completely unnecessary stress about the tournament in the final couple of weeks of production (including having to cut a round off the original projected 10 rounds), I had a lot of fun writing for this tournament and I hope everyone that played had a lot of fun playing it. Given the amount of new tossup answers and lead-in material introduced(?) at this tournament, I hope you maybe learned something too, even if all you learned was that the guy who yells "SCIENCE!" is Magnus Pyke.

[Mike Bentley]

What the hell is an S-Value in a binary tree?

[cvdwightw]

What the hell is an S-Value in a binary tree?

This is probably me being terrible with CS; it came up as a clue to "heap" in a Minnesota Undergraduate Tournament tossup, maybe that helps.

[Important Bird Area]

lots of ... "thinly disguised geography"

If this had actually been announced, I would have showed up to play.

[Skepticism and Animal Feed]

I imagine that many of the complaints about "stealth geography" were in regards to my "Hudson Bay" tossup. All I can say is that apart from Belcher Island and the impact crater clue, it was all stuff that came up in a glaciology class I took.

[Schweizerkas]

Thanks for making this event happen, Dwight et al. I really enjoyed playing on it. Speaking just for myself, I found the quirkier answers an amusing addition to the set. After listening to endless descriptions of random proteins and amino acids, it was nice to be able to get tossups on things like "hippos." And the S-Value tossup, although perhaps ungettable before the NAQT giveaway, still had the best answer of any tossup since RMPFest's epic tossup on Ray Luo.

Excuse me while I go study the product line of The Learning Company.

[BuzzerZen]

I wrote all the computer science for the tournament. I'd really appreciate feedback of any kind from anyone who cares to offer it.

[recfreq]

The computer science in this tourney seemed quite good. Almost all the stuff are on stuff that comes up regularly, and the questions all had interesting leadins and appropriate middle clues, although probably consistently a bit easier than the other categories. Now, for some constructive comments.

Although I was probably the only person at the West coast mirror capable of surmising STRIPS, the AI planning paradigm, I thought it was interesting to see as an answer (I had encountered it in a graduate AI course on knowledge representation 4yrs ago, and used it for a homework once, but alas could not pull it). I do wonder, however, whether the canon expansion could be accompanied in that field with something closer to what people would know, namely things like backpropagation, resolution, Bayes/decision nets, Viterbi/EM algorithm, reinforcement learning, or even a TU on planning.

The genetic algorithms question was rather easy to get early with or without real knowledge. I'd imagine many people powered it. First, it said "these algorithms of AI," then it was very clear it was talking about "creating variation by altering data at indiv loci," all in the power. Although I don't think GA is a huge part of AI (not taught at a large fraction of intro AI courses, or just touched on by name only), it would have been good to see clues like Holland's thm or two-armed bandit that are more on the theory side. I'm still not convinced GA has enough good CS clues to write on, though.

Other than that question, everything else was quite enjoyable to play on. I especially liked the Dijkstra TU. (Oh, and if ENIAC belongs to the CS distro, I thought that question'd benefit from less transparency, or at least not have the von Neumann clue in the power.)

[Auroni]

The tossup on "the Super Solver games" pretty much made my day. I had no idea that anyone else liked those games as much as I did, and I was happy to see them immortalized in tossup form (although, I was not too fond of the initial "Series or individual games acceptable". sentence, other examples of which did show up a bit in this set).

[cvdwightw]

(although, I was not too fond of the initial "Series or individual games acceptable". sentence, other examples of which did show up a bit in this set).

As far as I remember, there were only two "either X or Y is acceptable;" both were designed to improve accessibility. One was on the Super Solvers series, since I figured people were more likely to remember the names of Outnumbered! and Gizmos and Gadgets rather than the actual name of the series; the other was on that paper that Alan Sokal sent to Social Text that was basically a bunch of garbage about cultural relativism in physics or something, where I figured that there was no way in hell that people were going to pull the name of the paper but it didn't seem right to keep using "this author" when describing excerpts from the actual paper. If these questions somehow threw people off, I apologize.

[Mike Bentley]

The computer science in this tourney seemed quite good. Almost all the stuff are on stuff that comes up regularly, and the questions all had interesting leadins and appropriate middle clues, although probably consistently a bit easier than the other categories. Now, for some constructive comments.

Although I was probably the only person at the West coast mirror capable of surmising STRIPS, the AI planning paradigm, I thought it was interesting to see as an answer (I had encountered it in a graduate AI course on knowledge representation 4yrs ago, and used it for a homework once, but alas could not pull it). I do wonder, however, whether the canon expansion could be accompanied in that field with something closer to what people would know, namely things like backpropagation, resolution, Bayes/decision nets, Viterbi/EM algorithm, reinforcement learning, or even a TU on planning.

The genetic algorithms question was rather easy to get early with or without real knowledge. I'd imagine many people powered it. First, it said "these algorithms of AI," then it was very clear it was talking about "creating variation by altering data at indiv loci," all in the power. Although I don't think GA is a huge part of AI (not taught at a large fraction of intro AI courses, or just touched on by name only), it would have been good to see clues like Holland's thm or two-armed bandit that are more on the theory side. I'm still not convinced GA has enough good CS clues to write on, though.

Other than that question, everything else was quite enjoyable to play on. I especially liked the Dijkstra TU. (Oh, and if ENIAC belongs to the CS distro, I thought that question'd benefit from less transparency, or at least not have the von Neumann clue in the power.)

Eh, I thought the CS was a bit hit or miss. I'd say close to 50% of it I had figured out in the first one or two clues and then just waited for confirmation later in the questions. Some of this is obviously related to me having a degree in CS, but I felt some of it was also due to less than ideal question execution.

Dijikstra seemed like a good tossup, although I didn't see how it developed. The (Minimum Spanning) Trees tossup was also pretty good. The Church and TSP tossups also seemed pretty good. Towers of Hanoi seemed okay.

Less good were the tossups on Linux which pretty much mentioned it was an Operating System in the first clue and dared people to guess it would be the only OS approaching legit CS to ask about in a tossup. I seem to remember the Genetic Algorithms question starting with something like "these evolutionary algorithms" which was not a good idea. The Huffman Coding tossup made it clear that it was a coding method that used binary trees very early. The Four Color tossup looked pretty much impossible to power, and then dropped it's alternative name of Guthrie's Problem which is decently well known. The Regular Expressions tossup was pretty obvious it was going that way from the start. ENIAC was pretty much a first clue or nothing buzz, as it was made clear it was a general purpose early computer that's actually important in the first clue. I've never heard of STRIPS and am surprised Ray has heard of it. I guess it could be important, but I feel that would have been better served as being a tossup on something easier and AI-related.

Anyways, I still appreciate playing in a tournament where there is 1-2 CS questions per round, and overall the tournament was more fun than I thought it would be. Thanks for everyone who worked hard to write it.

[yoda4554]

the other was on that paper that Alan Sokal sent to Social Text that was basically a bunch of garbage about cultural relativism in physics or something, where I figured that there was no way in hell that people were going to pull the name of the paper but it didn't seem right to keep using "this author" when describing excerpts from the actual paper.

"Transgressing the Boundaries: Toward a Hermeneutics of Quantum Gravity"! This would have been the only question I would have gotten in the whole tournament, probably; I actually look forward to seeing how this would've worked. By the way, for future writers of such a question, "The Sokal Hoax" is the commonly-used shorthand for this.

[cvdwightw]

Just making sure that the mid-Atlantic either played this or has given up playing it before I send the whole thing to the packet archives. I'll send things as individual rounds instead of the big batch of questions, so people can download one round at a time.

[Theory Of The Leisure Flask]

Just making sure that the mid-Atlantic either played this or has given up playing it before I send the whole thing to the packet archives. I'll send things as individual rounds instead of the big batch of questions, so people can download one round at a time.

The Mid-Atlantic (i.e. ten round Chris Ray-Mehdi Razvi deathmatch) is finally going to play this at the MUT mirror on Saturday. (Don't worry, I'm not playing.)

[Cheynem]

We played some of the packets at practice and they were pretty entertaining even to a science dummy like me. I managed to get two tossups on the "Traveling Salesman Problem" and "Creation Science."

[cdcarter]

http://collegiate.quizbowlpackets.com/a ... scinon.zip