The Quizbowl Resource Center

It was recently brought to my attention that during the DII tiebreaker round, one of the half-packets that was used to break a tie had quite a few geography questions in it, which may have skewed the results of the half-round in favor of teams with stronger geography players. This got me to thinking that it might be a good idea to consider creating a more fair distribution for any half-packet tiebreakers that come up because such half-packets could be more science-heavy, geography-heavy, or whatever topic-heavy than usual. Of course the distribution tends to even out moreso over the course of whole packets as opposed to half-packets, but we know that sometimes there are more questions from some subject areas in the first half of a game relative to the second half, and that's natural, but I do think that special care should be taken to make the distribution more fair over the course of half-packets, especially when these half-packets decide the placement of teams on an annual basis.

Round 15, used for the playoff tiebreakers, had 1/1 geography, both in the first half.

bt_green_warbler wrote:Round 15, used for the playoff tiebreakers, had 1/1 geography, both in the first half.

I was the one who mentioned this to Jake. The packet that I mean is packet 8. I'm writing out the rest of my post now.

Round 8 indeed had two geography tossups and a part-geography miscellaneous tossup in the first 12 tossups.

bt_green_warbler wrote:Round 8 indeed had two geography tossups and a part-geography miscellaneous tossup in the first 12 tossups.

I don't have my notes on hand, but I recall the first half of this round being science heavy (so much so, it caused one of my players to tell me that he should sit out for the tiebreaker rather than play it as I intended for him to do). I'll double check this when I get back to my apartment later tonight.

EDIT: The fact that the third geography tossup was miscellaneous evens out the distribution a little, but still, why three geography tossups in half a packet?
EDIT: Ted Hughes is not Ted Williams and therefore is lit rather than sports.

I agree with you Charles, this packet seemed slightly heavy in science, given the distribution of the other two Big Three categories.

Now firstly, I understand that NAQT uses per-tournament distribution rather than a per-packet distribution. In my opinion this is kind of problematic, but that's neither here nor there. What bothers me about the distribution, however, was how this caused the half-packet tiebreaker that we played with Northwestern to be rather unrepresentative of NAQT's distribution. (I have no idea how the rest of packet 8 was distributed, because I left before I could watch Northwestern play WUSTL.) Since that game had the potential to knock us out of the top bracket (which it did), it seems to me that it should have had some resemblance to the overall distribution.

By multiplying the NAQT college set distribution by .92308 to adjust for the fact that Division I uses 24/24 packets instead of 26/26, and then dividing by half, the following is approximately what an "internally evenly distributed half" of a packet would have looked like:

2.15/2.1 Lit
2.25/2.2 Science
1.05/1.0 FA
2.25/2.2 History
0.6/0.6 Geography
0.4/0.35 Myth
0.75/0.7 Pop culture
0.3/0.25 Sports
0.55/0.55 Social Science
0.5/0.45 Theology/Philosophy
0.65/0.65 CE
0.5/0.9 Miscellaneous

Now, in our round 8 tiebreaker against Northwester, we played 12/11. According to my notes, that half-packet contained:

1/2 Lit (Dreser; Knight's Tale/Boccacio/Two Noble Kinsmen, Holy Sonnet X/Valediction: Forbidding Mourning/The Flea)
3/2 Science (alkynes, Einstein, d orbitals; e/algebraic/pi, phenol/BPA/triclosan)
1/0-1 Fine Arts* (all of which was music: Samuel Barber and maybe the jazz bonus)
1/3-4 History* (Charles Sumner; sit-in/Greensboro/CORE, USS Missouri/USS Yorktown/USS Indianapolis, Lloyd George/Atlee/Beveridge)
3/0 Geography (Bali, Sea of Azov, Guinea)
1/0 Myth (Maya)
1/1-2 Pop Culture* (someone named Monet; True Blood/Treen?/K-ville?, and maybe the jazz bonus
1/0 Sports (Ted Hughes)
0/0 Social Science
0/1 Theology/Philosophy (Zeno/Parmenides/arrow)
0/0-1 CE*
0/0 Miscellaneous

*The uncertainty in history/CE and FA/PC is due to my ignorance of whether the (Coke/Soros/Beck) bonus was classified as history or CE and whether the jazz bonus (Sonny Rollins/Davis/Saxophone Colossus) is classified as miscellaneous fine arts or as pop culture music.

I admit that it's possible that I'm misclassifying some of these myself- if some of the tossups that I identified as geography are actually history, I'd appreciate it if those could be posted. That would cause history to even out, and if the jazz bonus is classified as miscellaneous FA instead of miscellaneous pop culture, that would even out FA. (The sub-distribution of FA would still be out of whack, though, as 100% of FA would still be music and not visual.) In any case, Jeff, if you could post the missing bonus from the first half of round 8, that might clear up some things.

It's possible that the one missing bonus would even out the missing lit or the entirely absent social science. But by comparing these numbers to the distribution I recorded, it's clear that some categories were significantly over-represented (geography) and under-represented or totally absent (lit, and social science and miscellaneous, respectively), and that others, like history, (and geography, again) are just plain lopsided in terms of evenness of tossups/bonuses. History could not be helped by the missing bonus, since it was already either just 20% or 25% tossup and 80% or 75% bonus.

I'm not arguing that NAQT failed to live up to some sort of promise, since neither half-packets nor whole packets have any set distribution. But I feel like this is certainly a problem that NAQT should address the distribution of half-packets, since half games have to be used to settle many tiebreakers. No one wants to be knocked out of the top bracket because one's packet had a dearth of lit/history/social science and an excess of pop culture/sports/geography.

Papa's in the House wrote:

bt_green_warbler wrote:Round 8 indeed had two geography tossups and a part-geography miscellaneous tossup in the first 12 tossups.

I don't have my notes on hand, but I recall the first half of this round being science heavy (so much so, it caused one of my players to tell me that he should sit out for the tiebreaker rather than play it as I intended for him to do). I'll double check this when I get back to my apartment later tonight.

3/2 science in the first half out of 5/5 total.

Ben Cole wrote:Now firstly, I understand that NAQT uses per-tournament distribution rather than a per-packet distribution. In my opinion this is kind of problematic, but that's neither here nor there.

Just repeating once more for the record that we do balance each packet to the best of our ability (cf. this thread).

Ben Cole wrote:*The uncertainty in history/CE and FA/PC is due to my ignorance of whether the (Coke/Soros/Beck) bonus was classified as history or CE and whether the jazz bonus (Sonny Rollins/Davis/Saxophone Colossus) is classified as miscellaneous fine arts or as pop culture music.

Koch/Soros/Beck: CE
jazz bonus: FA

The Ted Hughes tossup was literature.

DII ICT round 8 wrote:After this man married his third wife, Carol Orchard, he discussed working on their farm in his 1970 ~Moortown Diary~. A year earlier, Assia Wevill, his second wife, had committed suicide. "A crush of diamonds, incredibly bright" was his poetic description of his first wife in (*) ~Birthday Letters~; that wife known for poems like "The Colossus" also committed suicide. For 10 points--name this husband of Sylvia Plath.

The twelfth bonus would have been another PC bonus:

For 10 points each--name these fashion trendsetters:

A. This Italian designer rose to fame after dressing Richard Gere in the film ~American Gigolo~; he also dressed Christian Bale for ~The Dark Knight~.

answer: Giorgio _Armani_

B. This London-born fashion-leader came up with the miniskirt in the 1960s and hot pants in the early 1970s.

answer: Mary _Quant_

C. In 1993 this American, born in the Dominican Republic, became the first U.S. citizen to design for a French couture house. He is renowned for opulent evening dresses.

answer: Oscar _de la Renta_ (or Oscar Aristides _Renta_ Fiallo)

Ben Cole wrote:1/0 Sports (Ted Hughes)

Fred wrote:

Ben Cole wrote:1/0 Sports (Ted Hughes)

Whoops, that's my fault, I confused him with Ted Williams because I wasn't paying attention very well.

Those better even out lit and fine arts, but the sub-distribution of fine arts in that half is still 100% music and 0% visual, instead of being about 50% visual as NAQT's distribution dictates. History is still lopsidedly 1/3, 75% of it was American, and unless I've mis-categorized some other question, social science is entirely absent.

I'm glad to know NAQT does balance packets, but it seems that there is still something to be desired about how tie-breaking half-packets are distributed.

Ben Cole wrote:it seems that there is still something to be desired about how tie-breaking half-packets are distributed.

This is certainly true; I have asked R. what we can do to improve within-packet distributions for next year's tournaments.

bt_green_warbler wrote:

Ben Cole wrote:it seems that there is still something to be desired about how tie-breaking half-packets are distributed.

This is certainly true; I have asked R. what we can do to improve within-packet distributions for next year's tournaments.

Wouldn't the simplest thing to do would be to have a standard per pack distribution? Is there a specific reason that NAQT does not do this? Excuse my ignorance, if this is an often-tread topic.

Bartleby wrote:

bt_green_warbler wrote:

Ben Cole wrote:it seems that there is still something to be desired about how tie-breaking half-packets are distributed.

This is certainly true; I have asked R. what we can do to improve within-packet distributions for next year's tournaments.

Wouldn't the simplest thing to do would be to have a standard per pack distribution? Is there a specific reason that NAQT does not do this? Excuse my ignorance, if this is an often-tread topic.

This wouldn't solve the problem (even with a standard per packet distribution you can still put four science tossups in the first half; given even an inconsistent by the packet per-tournament distribution, you can still put half of all the [broad category] in the first half.

The reason NAQT goes per tournament is they'd like to put explicit minima and maxima on, say, Norse myth per tournament. Let's say they want 4/4 per tournament--they can't put per-packet requirements on that, since they don't want "1/0 or 0/1 per packet" even. This is actually similar to what ACF does when RMP is given only 2/2 (i.e. often): if you considered religion individually a top-level category, you'd say "why is this per-tournament distributed rather than per-packet distributed."

Crazy Andy Watkins wrote:

Bartleby wrote:

bt_green_warbler wrote:

Ben Cole wrote:it seems that there is still something to be desired about how tie-breaking half-packets are distributed.

This is certainly true; I have asked R. what we can do to improve within-packet distributions for next year's tournaments.

Wouldn't the simplest thing to do would be to have a standard per pack distribution? Is there a specific reason that NAQT does not do this? Excuse my ignorance, if this is an often-tread topic.

This wouldn't solve the problem (even with a standard per packet distribution you can still put four science tossups in the first half; given even an inconsistent by the packet per-tournament distribution, you can still put half of all the [broad category] in the first half.

The reason NAQT goes per tournament is they'd like to put explicit minima and maxima on, say, Norse myth per tournament. Let's say they want 4/4 per tournament--they can't put per-packet requirements on that, since they don't want "1/0 or 0/1 per packet" even. This is actually similar to what ACF does when RMP is given only 2/2 (i.e. often): if you considered religion individually a top-level category, you'd say "why is this per-tournament distributed rather than per-packet distributed."

That makes enough sense. I was conceptualizing the idea of "within-packet distributions" as related to 26/26, and not, say the 13/13 (or 12/12, in DII) that would result from a half-game tiebreaker.

Given that NAQT will know in advance of the tournament which packets will be reserved specifically for tiebreaking (i.e. packet 8 this year), could there be extra attention paid to the half-wise distribution of the questions that go in that packet?

styxman wrote:Given that NAQT will know in advance of the tournament which packets will be reserved specifically for tiebreaking (i.e. packet 8 this year), could there be extra attention paid to the half-wise distribution of the questions that go in that packet?

Interestingly, the second most likely packet to be involved in a half-packet tiebreaker is Packet 15, the same packet that is the most likely overall to be read to completion.