Power Matching

[Stained Diviner]

If anybody has comments, questions, criticisms, or suggestions on power matching (or card systems), here is a good place. Feel free to comment on HSNCT specifically or power matching in general.

For some info on how the HSNCT cards played out on Saturday, feel free to look here. You can also look here. The two files currently have the same information. Only I can edit the first one, while anybody can edit the second one. If you are going to edit the second one, study the system before you do anything, look up initial cards here, and for the love of all that is holy do not sort anything ever. Of course, all of the info is already at naqt.com, so only do this if you have absolutely nothing else to do with your time.

There is already one suggestion made here from Chris Chiego. He or anybody else is free to go into details, but I will give a few warnings on this. One warning is that NAQT already produces 27 packets for HSNCT, and it is difficult to add to the number of packets without making the questions more difficult because there are only so many good answers to use in a high school tournament. Also, the double elimination system on Sunday is pretty close to maxed out for 10 rounds, so allowing more teams to play on Sunday probably means another round of questions and another half hour of the tournament. In addition to that, you just get into general questions as to when teams should be given second chances and when they shouldn't be.

[Sniper, No Sniping!]

How many games that featured two teams with different records (coming into the game) were played Saturday?

[Stained Diviner]

I don't know how many actually occurred. Before the tournament, we calculated an expected value of 38 such matches, with a range anywhere from 4 to 75.

[cchiego]

Why not do something better than just the meaningless consolation rounds on Sunday? Why not have a way to differentiate between the 5-5 and 6-5 (first round playoff loss) teams with 2 or 3 more rounds? Perhaps they can't be part of the double elim playoffs, but between 10 AM and 12 PM it seems like playing 2-3 more matches would be eminently possible to provide better final placement between the 5-5 and 6-5 teams.

It's kind of like a 2nd-tier playoff. The top teams after the first couple of rounds on Sunday are still in competition for the championship while the teams just a tier below are in competition for better placement.

This requires no new packets and no new staffers. It would simply take some staffers away from the current pointless consolation rounds and make Sunday more meaningful for many more teams.

It would not help correct for the way teams are currently penalized under the card system for winning early with much tougher matchups in the later rounds (see: Chattahoochee last year, Glasgow, Farragut, and Bloomington this year), but it would at least provide teams in the 5-5 range some way of differentiation and more meaningful matches.

[Maury Island incident]

I second Chris' idea.
My team, Lexington, went 5-5, but had a fairly tough schedule, and feel like we could have had a better placement than we did.
We won our consolation tournament pretty handily (although Bloomington was not at full-strength), but would much rather have played in matches like Chris proposed. One of our team members even changed his flight to fly out Saturday night because he did not want to play consolation matches.There are many 6-5 teams who we probably could have beaten, and under that system, we would have had the chance to beat some of those teams and improve our final standing.

[Stained Diviner]

I hope that somebody who works on the consolation rounds will comment on your suggestion--I don't have enough experience in that area to determine the feasibility of using the consolation rounds to determine a 5-5 and/or 6-5 consolation champ.

As far as the four teams Chris mentioned, they all follow a similar pattern. They started the day strongly, which meant that by the middle of the day they were facing elite teams that they lost to. At the end of the day, they (with the exception of Farragut) faced teams with worse stats than their own but lost to those teams. The fact that they started the day strong meant that they played a strong schedule overall. The fact that they lost to teams with worst stats than their own meant that they ended up with a record weaker than their stats.

When the card system works well, teams face opponents with the same record, which is what happened in all four of those cases. As much as possible, the cards match a team coming off a loss with a team coming off a win. This is because teams coming off a loss tend to be the better teams with a given record, while the opposite is true for teams coming off a win. This didn't work well for Farragut, since Northmont A was coming off a win but was actually one of the strongest (if not the strongest) 6-3 teams. The other teams Chris mentioned, as I said, lost to opponents with worse stats than their own.

Lexington's case is a little different, since they stayed around .500 all day. They had to win their last two matches, which the card system makes difficult. Throughout the day, they tended to get opponents that leaned a bit towards the difficult side without really being outliers, and they took a close loss at the end of the day that eliminated them.

[Maury Island incident]

Yes, the teams we played were not that difficult, and we played a mostly reasonable schedule.
The one time when the card system gave us problems was when we faced Buchholz. They had card 116 at the time, after they lost to LASA A. One of our low card games was essentially taken away. The other teams we lost to were all had pretty reasonable cards for their rank at that point, though we did face six playoff teams, and four trophy teams.

[AKKOLADE]

Does the card system account for the strength of teams with those records (PPG or whatever)? Is there anyway it could?

[Stained Diviner]

The cards do not account for statistics during the tournament, though as I said above they are set up so that a team that is expected to be from the top half of the teams with a given record is scheduled to play a team that is expected to be from the bottom half.

The best example of how this works is seen in the matchups of the 1-1 teams. If seeding is correct and there are no upsets, then the top half of teams win their first match, and the bottom half of teams lose their first match. After the winners play winners and losers play losers in the 2nd matches, half of the teams are 1-1. Of that half, half of the teams won their first match and lost their second match, and are expected to be in the upper half of the 1-1 teams. Half of the teams lost their first match and won their second match, and are expected to be in the lower half of 1-1 teams. So, the teams that won then lost play the teams that lost then won. If seeding is correct and the better team wins every time, then the teams that won then lost their first two matches should always beat the teams that lost then won their first two matches. In reality, the WL teams usually beat the LW teams, but there are plenty of exceptions. The same principle is used for most matches in the tournament, though it's imperfect because winning records have more teams coming off wins than losses, and losing records have more teams coming off losses than wins.

The teams are seeded so that the elite teams don't play each other in the morning, but the seeds are purposely hidden so that no team is labelled as the 272nd best team in the tournament before it begins.

[gustavus.adolphus]

The teams are seeded so that the elite teams don't play each other in the morning, but the seeds are purposely hidden so that no team is labelled as the 272nd best team in the tournament before it begins.

How do you think that philosophy differs/is superior to NHBB's releasing all the seedings before the tournament?

[Stained Diviner]

(I'm going to ignore Adam's question because, though I agree with NAQT's decision, that is not my decision to make, and it would be better if a NAQT Member explained their reasoning. This post is a further reply to Fred's question.)

As far as whether PPB or PP20H could be used during a tournament, there are two ways this can be done:

1) Many years ago, NAQT used to hold up the tournament at a few points in the afternoon to get stats entered and match up the strongest team with each record with the weakest team with each record. This practice was eventually dropped because teams complained about the half hour wait between each round to compile stats. Such a system is now impossible with the way the byes are set up, since some teams need to start their eighth match at the same time that other teams are starting their seventh match, making it impossible to get all the stats together. Also, NAQT has 9 people entering stats during the tournament and other people looking at the scoresheets when they come in to make sure all the numbers are in there, and it still takes a lot of time to enter all the matches, since there are 85 matches each round and total of 1360 matches.

2) The schedule is currently set up so that the 5-4 teams have a Round 15 bye and play each other on the first floor during Round 16. If every match room had a computer that allowed results entry, and it was reliable, it would be possible to get those particular matchups done quickly so that they are always the best 5-4 team playing the worst while avoiding repeat matchups. I would like to see this happen at some point, but my guess is that it is still a few years off. With byes and rooms that are spread out over a somewhat wide area, it would probably be a mistake to do that sort of thing on a tournament-wide basis for every match.

[cvdwightw]

I haven't posted anything in a while, but I wrote some stuff for HSNCT, so I thought I'd check out the resulting discussions.

It is theoretically possible, but perhaps not plausible in practice, to run a card system that seeds 6-3 and 5-4 teams for their last game. Let me explain using the 256 team power match.

After each team has played 8 games, there is 1 8-0 team, there are 8 7-1 teams, 28 6-2 teams, 56 5-3 teams, 70 4-4 teams, 56 3-5 teams, and so on down to 1 0-8 team. This necessitates that in the 9th game played by each team, there are 8 games between teams with different records.

After the ninth game, there are 22 teams with at least 7 wins. Of the "playoff interest" teams, there will be 2 teams at (7-2 or 6-3), 40 teams at 6-3, 2 teams at (6-3 or 5-4), 61 teams at 5-4, and 2 teams at (5-4 or 4-5). In other words, we could have anywhere from 40 to 44 teams at 6-3 (and thus playing to get into the winner's bracket) and anywhere from 61 to 65 teams at 5-4 (and thus playing to get into the the playoffs). Note that in the majority of these cases we will have at least one 5-4 vs. 4-5 matchup.

With 84 game rooms (not an unreasonable assumption - this year had 85, I think), you would need until Round 13 for each team to have completed their eighth game. A maximum of 136 teams will play their 9th game in Round 13. All of these 136 would be on bye in Round 14, so after Round 14 all teams have played 9 games. You would need all 107 teams of "playoff interest" to be part of that 136. You could then work the schedule such that those 107 teams had another bye in Round 15.

On those particular 107 cards, you would need to indicate a place to meet (one place for 6-3 and one place for 5-4 teams). You would then need to (1) collect all cards from teams at that particular meeting place; (2) hope that your statkeepers can get through those "high-priority" Round 13 games by the end of Round 15; and (3) re-distribute the cards at the meeting place.

Herein lies the problem: in the vast majority of cases, you will have either an odd number of 6-3 teams, an odd number of 5-4 teams, or both. I guess at that point you'd have either (1) the bottom 6-3 team play a 7-2 team; (2) the bottom 6-3 team play the top 5-4 team; (3) the top 5-4 team play a random 4-5 team; or (4) some combination thereof.

In short: it can be done, but you need (a) really fast statkeepers and (b) two highly competent people in control rooms whose jobs in Rounds 13-16 boil down to "crowd control and handing out cards really quickly."

[Stained Diviner]

Dwight's suggestion adds some inefficiencies. With 256 teams playing 10 matches in 16 rounds, you need 80 rooms. Adding more rooms makes the tournament less efficient in terms of requiring more moderators and rooms than necessary. His system also probably requires additional stats peoples so that the necessary stats are done on time. Also, with 272 teams, the uncertainty as to the number of 6-3 and 5-4 teams goes up.

It would take me some time to figure out whether the schedule he describes for the last few rounds is feasible, and how many rooms it would actually require. Another issue is that NAQT arranges the bubble matches so that they are in rooms with moderators with a proven track record, and that would be more difficult if there were a large number of such matches going on in the same round.

A minor quibble is that you would not use cards to send teams to their last match--you would post it in person and electronically.

[Important Bird Area]

The teams are seeded so that the elite teams don't play each other in the morning, but the seeds are purposely hidden so that no team is labelled as the 272nd best team in the tournament before it begins.

How do you think that philosophy differs/is superior to NHBB's releasing all the seedings before the tournament?

We think the potential cost in hurt feelings/public argument will outweigh the potential small benefit in improved rankings.

2011 HSNCT seeding discussion

[Important Bird Area]

Why not have a way to differentiate between the 5-5 and 6-5 (first round playoff loss) teams with 2 or 3 more rounds? Perhaps they can't be part of the double elim playoffs, but between 10 AM and 12 PM it seems like playing 2-3 more matches would be eminently possible to provide better final placement between the 5-5 and 6-5 teams.

We believe that many teams would be unwilling to stick around for such tiebreaker games. Better to differentiate 5-5 teams via some measure of quizbowl skill (pptuh) rather than differentiate by "which team forfeited to go see the Art Institute."

[Fado Alexandrino]

Official placement games for non playoff and early 6-4 playoff losses also solves the problem of disorganized consolation rounds, even if it is only for 3-4 rounds. The tournament ending at 11 am should give the teams that aren't in the top 33 or whatever plenty of time to see the city on Sunday.

[Great Bustard]

The teams are seeded so that the elite teams don't play each other in the morning, but the seeds are purposely hidden so that no team is labelled as the 272nd best team in the tournament before it begins.

How do you think that philosophy differs/is superior to NHBB's releasing all the seedings before the tournament?

To clarify here, NHBB does not release all the seedings before the tournament. We only release seeds for roughly the top 1/3 of teams in advance of the event, while teams who were seeded in the top half can see that in the file showing comparisons of 1-3 seeds to performance. Some teams who were seeded in the bottom half but who made the Upper Bracket rounds (i.e. top 3 in their morning group of 6) can figure out their seed from the same file, but that's it.
NHBB does not release seedings of teams in the lower parts of the draw for precisely the reason that we don't want teams tagged with the "6th seed in the group" moniker. We release the seedings for the higher teams because I think transparency is a good thing up to a point, and teams like to see where they stack up.

[cchiego]

We believe that many teams would be unwilling to stick around for such tiebreaker games. Better to differentiate 5-5 teams via some measure of quizbowl skill (pptuh) rather than differentiate by "which team forfeited to go see the Art Institute."

I don't think so and I hope that NAQT solicits more feedback on this from teams. 3 more competitive games over 4 rounds from 9 AM to 11 AM would be just the right amount of Sunday morning competition to better rank teams than the very much opponent-dependent PPTUH and a far better use of time/staffers than consolation rounds. Note that the NSC does Sunday morning games for all teams and manages to do fine. Besides, with no buzzers ready until 12:30-1 PM anyways, many teams are sticking around regardless.