ICT Seeding

[Cheynem]

Since this has nothing to do with questions, it seemed silly to put it in the private forum, but if it's more appropriate to group topics together, feel free to move it.

Every year the seeding for ICT's prelim brackets seems to rankle people. My point here is not to complain about "one bracket being stronger than the others," it's to offer up two comments.

1. Were the "seeds" so to speak for the prelim bracket reflected in the final game of the prelims (i.e. was seed #1 supposed to be playing seed #2?). If so, there were some crazy things about that that seemed to reflect an over-reliance on D-value/SCT results and not enough on actual information about who was playing (i.e., Illinois B had a good SCT performance and D-value but had nothing like the same lineup; WUSTL seemed to have a very low seed despite having Gordon and Charles).

2. Some of the brackets ended up very annoyingly regional. In Minnesota's prelim bracket, we were pitted against Michigan, Illinois, Ohio State, and Carnegie Mellon. We had played Michigan, Illinois, and Ohio State at countless tournaments before ICT and in fact had already played Carnegie Mellon at SCT. Coming out to a national tournament and then playing about half of your prelim games against teams from your region or that you had just played is kind of annoying.

In general I thought the prelim brackets turned out to be balanced okay. The number of logjams for the last playoff spot reflects the extremely strong field this tournament had--several of the second bracket teams were top bracket contenders.

[bt_green_warbler]

an over-reliance on D-value/SCT results and not enough on actual information about who was playing

...

Some of the brackets ended up very annoyingly regional.

Guilty as charged on both counts. I should have taken more time to review these.

[ValenciaQBowl]

I suppose having Delaware be a surprise to most observers was the culprit, but the Div II bracket including Harvard, Chicago, Delaware, and Chipola seems pretty rough in retrospect.

[Black-throated Antshrike]

For nothing more than curiosity, could you tell me what Delaware was actually ranked?

[cchiego]

There are a number of issues with the current tiebreak and seeding procedures that could stand to be improved.

First the tiebreaks: The NAQT distribution is already skewed from being a tournament-wide distribution rather than a by-packet distribution. Playing only 12 and 12 increases the chances of an even more skewed distribution when deciding who makes the various brackets. Given the vagaries of the initial seeding and the immense importance in terms of final rankings of the playoff brackets, it makes sense to make sure that these tiebreaks are done to maximize rather than reduce the likelihood of the better team winning. Thus, tiebreaks should be full games rather than half-games.

This is eminently feasible from a logistics standpoint. Playing two full timed matches rather than two untimed half matches isn't much more of an increase in time. You put the best readers in the tiebreak games and you get done with 2 full tiebreak matches in 45 minutes. ICT already ends plenty early; the only drawbacks are the other teams have a slightly longer lunch (and that whole lunch break could stand to be shortened).

If you do keep the half games, they should be on special, balanced packets designed specifically with the knowledge that they'll be used for tiebreak games--and this isn't a terrible idea either for full games on tiebreak packets either.

Seeding is harder to address because of the lineup changes, host teams not playing, and the significant increase in difficulty from SCT to ICT. Is there any way to have teams play 3 optional "seeding" matches on Friday night like they used to do at the old Vandy tournaments? It would be another source of information that all teams could have the option of submitting to help improve the accuracy of the seeding. Players and staffers are already there, it would take less than 2 hours to run, plus there always seem to be extra packets.

Upsets are fine and will happen in any tournament. But there's no reason that a national championship with a manageable number of teams should be so strongly dependent on the vagaries of initial seeding based on needlessly limited information and tiebreak procedures that favor upsets.

[Mechanical Beasts]

First the tiebreaks: The NAQT distribution is already skewed from being a tournament-wide distribution rather than a by-packet distribution.

I don't think this is much more true than it's true that the ACF distribution can "skew" if the chemistry question is on the bio-y side and the physics tossup is on the pchem-y side. Packet-to-packet variance is universal and the tournament-wide distribution actually does pretty little to change that. Though I agree with your later conclusion: I think tiebreaks should be full games.

That said, I think that adding four packets to the set--because if there are extra packets, then they were needed for a plausible playoff scenario or for a source of tiebreaker questions (or both!)--that the fewest writers are qualified (or even cleared) to write for is a tall order.

[The Quest for the Historical Mukherjesus]

I certainly had some problems with the seeding of this tournament, but given the sheer number of contending teams at this years' ICT, I don't really think I could have asked for anything better.

[Magister Ludi]

I'm sure people will be mad about the seeding of ICT, but I think any problems were a result of a uniquely even field rather than anything NAQT did wrong. I know I was initially angry about Harvard's seeding as the third number two seed, and was specifically irked we were below Minnesota because we were the reigning champs and only lost our leading scorer while Minnesota finished second and lost two of their top three scorers. But we finished seventh (technically tied for sixth) and lost to Minnesota, so it seems that NAQT's seedings were surprisingly astute.

I think teams are just angry about seedings because making the top bracket came down to one very close game between two evenly matched teams in most prelim brackets, so one or two buzzer races seem to hold an inordinate amount of weight in determining a team's overall placement. But that volatility is just inherent in the NAQT format. For example, UVA nearly didn't even make the top bracket when we beat them in the first half of a tie-breaker and their second mini-match against UCSD came down to the last question. Unless NAQT completely restructures their format, I don't know what kind of changes they could make to pacify teams who felt short-changed by their initial seeding.

[bt_green_warbler]

If you do keep the half games, they should be on special, balanced packets designed specifically with the knowledge that they'll be used for tiebreak games

This is an idea that I fully support, and I will make this happen for the 2013 ICT.

Full-game instead of half-game tiebreakers might be able to happen for certain tiebreak situations (we'll take a look at this).

Prelim seeding matches certainly won't happen (we won't write extra packets for something that's "optional"; and because it's "optional" I think it would be of limited utility anyway)

Keep in mind that we actually need backup packets; this year's final had a very straightforward resolution, and we still used 15 of the 18 packets. (where the hypotheticals are: 16 needed to resolve second-place ties, 17 if the trailing team had won the first game of the advantaged final, 18 in case of moderator error).

[Matt Weiner]

At some point, you have to win games against really good teams to advance in the tournament, regardless of what the format is, so I'm not hugely concerned with the fact that seeding may have prevented Team X from making the top bracket. With that said, the only real solution if a similar number of contenders is expected in future years is to expand the top bracket to 12 teams (the top 3 from each prelim) and correspondingly play 9 playoff games instead of 6; this would add about 90 minutes to the tournament in order to gain some additional amount of certainty about seeding and format fairness.

[setht]

At some point, you have to win games against really good teams to advance in the tournament, regardless of what the format is, so I'm not hugely concerned with the fact that seeding may have prevented Team X from making the top bracket. With that said, the only real solution if a similar number of contenders is expected in future years is to expand the top bracket to 12 teams (the top 3 from each prelim) and correspondingly play 9 playoff games instead of 6; this would add about 90 minutes to the tournament in order to gain some additional amount of certainty about seeding and format fairness.

I haven't really thought about this much, but I prefer Matt's solution to the alternative of stretching out tiebreaker games. His solution would also add 3 packets to the set; as I've said over here, I'm worried about the prospect of producing an 18 packet set next year, but if circuit involvement jumps way up, then presumably it would be possible (from a set production viewpoint) to switch to a 21 packet set and the extended schedule. I wouldn't mind adding 90 minutes to the tournament, but perhaps some players or staffers do.

-Seth

[DumbJaques]

I'd absolutely love to add some games to ICT, though clearly it's going to be incumbent upon us as a community to pitch in and write some more questions. As the tournament's already just a one-day affair, I think it could easily stand the addition of 90+ minutes to the schedule.