12 packets. 24 teams. Stacked field.

[jonah]

For the tournament I'm co-TDing on December 18, we have 12 packets and 24 teams (in the upper division). The field is, in my opinion, unusually strong for a regular-season tournament. I've come up with a proposed format that might be better than the obvious 4-brackets-of-6-then-rebracket-to-the-same in terms of having more matches between the top teams and eliminating paper tiebreakers. I am interested in opinions on this proposal.

Start with 6 brackets of 4. That's 3 packets. We'll call this the prelim phase.

Break ties in brackets. Now we've used 4 packets. For reasons that will become apparent later, we tell the teams not to talk about the tiebreaker match's questions.

Give the label 1 to the team that has the highest PPB among all teams that are top in their brackets. Give the label 2 to the team that has the second-highest PPB among all teams that are top in their brackets. And so on through 6.
Give the label 7 to the second-place team in the same bracket as 1. Continue giving the label n+6 to the second-place team in the bracket containing team n.
Give the label 13 to the team that has the highest PPB among all teams that are #3 in their brackets. Give the label 14 to the team that has the second-highest PPB among all teams that are #3 in their brackets. And so on through 18.
Give the label 19 to the last-place team in the same bracket as 13. Continue giving the label n+6 to the fourth-place team in the bracket containing team n.

Form two pairs of parallel playoff brackets: one containing 1,4,5,7,10,11 which is parallel to a bracket containing 2,3,6,8,9,12. Another pair of parallel brackets has 13,16,17,19,22,23 parallel to 14,15,18,20,21,24. Note that each of these contains three pairs of teams from the same prelim bracket.
Do a RR in these brackets, counting the already-played match between each pair of teams from the same prelim bracket. Now we've used 8 packets. Call this the parallel-brackets phase.

Now divide each bracket into a top half of 3 and a bottom half of 3. The top half of each bracket plays the top half of its parallel bracket, and the bottom half of each plays the bottom half of its parallel bracket. That's 3 more packets, so we've used 11. Call it the superplayoffs.

At the end, we might need to do a finals in the top superplayoff bracket (the one consisting of the top halves of the top pair of parallel brackets). We only have one packet for it, unfortunately, unless both teams involved in it didn't play the tiebreaker round. In that case, we can do a normal advantaged final. If not, we might be a little screwed and have to use half-packets, or something like that.

There are two potential problems I see with this: timeliness of stats (since we'd be doing the first rebracket before lunch) and the finals situation. I'm not terribly concerned about timeliness of stats, since we have a very good statskeeper with the ability to add another if we need to, no rebounds, and probably not a whole lot of subbing going on.

Feedback? Improvements? Accusations of insanity? I'm not invested in doing this format or anything, I'm just contemplating it as a possibility.

[Mechanical Beasts]

For any rebracket that has to happen prelunch, have the brackets drawn on a chalkboard in the control room (or paper in front of the statkeeper) and have the returning mods tally their match as a win or loss for each team when they report in. That way you can fall behind in stats (not like that's especially likely, but you know) and still have instant standings and knowledge of tiebreakers.

[Stained Diviner]

Andrew is correct.

One issue I have with Jonah's system is that it splits the field into a top half and bottom half after three matches.

To give teams a chance to cross over, we could seed teams 13-24 after the prelims the same way we are seeding 7-12, based on the stats of the top team in their prelim bracket. This would allow us to take the teams that finished first in the bottom half brackets, compare them to the teams that finished last in the top half brackets, and swap them for the last three rounds if a bottom half pool winner puts up better PPB than its corresponding top half pool loser. I'm not sure exactly how we'd treat records up to that point, but the teams that swapped would play three matches against their new level of opponent and we're talking about at best 7th place anyways.

[SoLegit12]

Why PPB? Why not use PPG? It aggregates tossup and bonus abilities to be more accurate across a tournament.
I guess it goes to the whole thing in the rankings forum, but I think for a single tournament PPG would be better

[jonah]

Why PPB? Why not use PPG? It aggregates tossup and bonus abilities to be more accurate across a tournament.
I guess it goes to the whole thing in the rankings forum, but I think for a single tournament PPG would be better

For cross-bracket comparisons, especially assuming imperfect seeding, PPB makes more sense because it's opponent-independent. Indeed, the teams being compared in this case have no common opponents.

[SoLegit12]

Why PPB? Why not use PPG? It aggregates tossup and bonus abilities to be more accurate across a tournament.
I guess it goes to the whole thing in the rankings forum, but I think for a single tournament PPG would be better

For cross-bracket comparisons, especially assuming imperfect seeding, PPB makes more sense because it's opponent-independent. Indeed, the teams being compared in this case have no common opponents.

I always forget quizbowl isn't a utopia of perfect seeding.
However, PPB can be extremely close. What's to say a team with 19.67 PPB is better than a team with 19.23 PPB? Maybe a team misheard a part or something. Perhaps when PPB becomes close (like <.5 ppb) PPG could be used to break the team? I am not an expert, but it seems good to me.

Edit: Also --> always

[Irreligion in Bangladesh]

Why PPB? Why not use PPG? It aggregates tossup and bonus abilities to be more accurate across a tournament.
I guess it goes to the whole thing in the rankings forum, but I think for a single tournament PPG would be better

For cross-bracket comparisons, especially assuming imperfect seeding, PPB makes more sense because it's opponent-independent. Indeed, the teams being compared in this case have no common opponents.

I always forget quizbowl isn't a utopia of perfect seeding.
However, PPB can be extremely close. What's to say a team with 19.67 PPB is better than a team with 19.23 PPB? Maybe a team misheard a part or something. Perhaps when PPB becomes close (like <.5 ppb) PPG could be used to break the team? I am not an expert, but it seems good to me.

Edit: Also --> always

If you're using PPG as a secondary stat to PPB - that is, teams are close in PPB so you're gonna look at PPG next - then what's going to make the difference is number of tossups converted. (It's also a big factor in PPG as a primary stat, of course.) Tossup conversion is dependent on how good of a team you are, sure, but it's also really, really dependent on how good your opponents are. That variable confounds things enough to hinder PPG as a ranking measure in the scope of "which team makes which playoff bracket" decisions.

[Aaron Goldfein]

I like this format a lot better than than the four pools of six idea. I feel like you'll almost certainly have two packets available for finals as the top two teams in the tournament will probably have no trouble finishing 1st or 2nd in their morning pools without tiebreaker rounds. With that in mind, I would only use the tiebreaker packet then if it's to distinguish between being 1st/2nd and being 3rd/4th. That is, if teams are tied for 1st/2nd by games, 1st should probably just be declared via PPG/PPB.

Also, what if teams are tied after the parallel brackets rounds? Do you just split them by statistical tiebreakers?

[Sen. Estes Kefauver (D-TN)]

I actually think there are many good reasons why you should not use points per bonus as a tiebreaker.

1) Your ability to answer bonuses may not be affected by your opponents, but the actual bonus conversion number you get is heavily dependent on the teams you play, because the number of bonuses you get to hear, and which bonuses they end up being, is just as dependent on your opponent's strength as your tossup conversion data.

2) Bonuses are very chaotic. There is no such thing as a set that perfectly calibrates the bonuses. It is impossible to ensure that hard parts are equally hard and the like. Because of the reliance on getting tossups referenced in 1), bonuses become even more chaotic because it becomes essentially impossible for the teams you are comparing to have actually heard all of the same bonuses. You will be using bonus conversion to compare teams who heard very different numbers of bonuses, teams who heard very different subject distributions of bonuses, and undoubtedly you will be using it to compare teams who didn't even play a good number of the same bonuses. Every little addition to the chaos mentioned above can make a team's bonus conversion fluctuate, and I don't feel that it is right for one team to win a tiebreaker because they happened to have heard a bonus with parts "Mark Twain-Tom Sawyer-St. Petersburg" while another team they're being compared to has a slightly lower bonus conversion because instead of a chance to hear that bonus, they got something like "Clifford Geertz-Margaret Meade-Samoa." Contrast all these problems with more tossup based statistics, where every team in the field has heard all of the exact same tossups in all the same rounds.

3) Bonus conversion is the easiest tiebreaker to get wrong. If you accidentally hit the 2 instead of the 1 key when entering a team's tossups, now suddenly all of your bonus conversion information for that team is inaccurate. Doing it by hand is not an option unless you literally have every staffer in a tournament working to run a tally from the scoresheets. Having such an important thing as a tiebreaker that gives a team a chance to win a tournament or not be so reliant on something that could go wrong in a heartbeat seems to me to be unacceptable.

[Stained Diviner]

If we use my change, then PPB is only being used on the six teams that won their preliminary brackets, and it is only being used at the end of Round 3 to make pools for the middle part of the tournament. The top seeds will almost certainly have heard 30+ bonuses out of the 60, since you can't win a bracket and not hear a significant number of bonuses.

Nobody is claiming that PPB is a perfect statistic, because it obviously is not. It doesn't take into account tossup answering skills, teams hear different bonuses, and, as Charlie stated, errors in keeping the statistic are easy to make and not as obvious when they occur.

However, having a statistic that does not depend on the strength of competition is very important. There is no such thing as an equally balanced schedule--in fact, there is not even an attempt to create an equally balanced schedule. Under our system, if you're the #1 seed, you will face in the first three matches is the #12 seed, and the weakest they will face is the #24 seed, with the three seeds of your opponent adding up to 49. If you're the #6 seed, the strongest opponent you face is the #7 seed, and the weakest is the #19 seed, with the three seeds adding up to 44. If you're the #24 seed, the strongest opponent you face is the #1 seed, and the weakest is the #13 seed, with the three seeds adding up to 26. In other words, the strengths of schedules are very uneven and are determined by performances at prior tournaments when you may have had a different lineup and by the judgement of the TD, which is often based on limited data. Furthermore, even if the TD judges your team appropriately, the strength of your schedule can be off if the TD judges the strengths of other teams incorrectly.

In other words, it is not fair to use PPG as a statistic. If we use it, then when we give a team a lower seed we are not only making it harder for that team to win its pool, we are also making it harder for the team to get a high seed even if it does win its pool, and even if its play is superior to that of other pool winners. If we use PPB, then teams that win will be judged in a way that does not reflect the fact that they have not played schedules of equal strength.

[Maxwell Sniffingwell]

Two problems with this, one minor and one probably a bit misguided:

a) You're requiring teams not to talk about the tiebreaker round - if more than 2 or 4 teams play in this round, that would be tricky. Players love to talk about questions at tournaments, and it would be tough to remember which round was the secret round later in the day. Effectively, you might need to enforce a blanket "please don't talk about any of the questions until the tournament is over" policy.

b) If I understand your scheme properly, it makes the matches at the start of the day pretty strong in determining finish. For instance, a team that takes two losses in the prelim phase - or even goes 2-1 but loses in the tiebreaker phase - can finish, at best, 13th. That means that a team could go 9-1 plus a tiebreak loss... and finish in the bottom half of the field. (2-1 in prelims; loss in tiebreak; beats 16, 17, 22, and 23 (already played 19); beats 14, 15, 18) That would make seeding correctly very, very important - for instance, if your 10 seed is three spots stronger than expected, your 7 seed could have that 9-1-but-13th finish.

[jonpin]

I find all this to be a solution to a problem that doesn't exist. 4x6 to 4x6 isn't the best way to do this. I guess you don't quite have enough to do 4x6 to 3x8 (that uses 11 packets for games, leaving none for tiebreakers and one for finals). Perhaps what you could do is 4x6 (1-5; TB 6); 3 tiers of 2x4 (7-9; TB 10). Semifinals (11) and finals (12). Or, if you wanted, and you didn't have to use tiebreakers earlier, you could have a best-of-three finals/3rd place/5th place on 10-11-12.

[jonah]

Also, what if teams are tied after the parallel brackets rounds? Do you just split them by statistical tiebreakers?

Yes, but this doesn't matter much. Other than a 5-way circle of death, which is just so darn unpleasant I'm not going to think about it, the only tiebreaker necessary would be between 3rd and 4th place in each of those brackets, in which case you're not in contention to win the tournament anyway so I don't feel terrible about the paper tiebreaker.

[Stained Diviner]

Jon, we would rather avoid ending the day with single elimination or only taking one team from the top pools (which is similar to single elimination).

Greg, I'm on the fence about your first objection--you may be right about that. Your second objection is why I made my modification, allowing for one or two teams to move from the bottom half to the top half if their stats justify it. If you use 4x6 to 4x6, then you probably get a 9-1 team finishing 7th, which is a similar problem though not quite as bad.

[jonpin]

Yeah I understand the desire to avoid SE or taking one team, though at some point, you have to say "there aren't enough packets". If you're going to insist on two finals packets and playing off ties, and given that I think it's a bad idea to have anything less than 5 prelim rounds (you don't want a mis-seeded team to go 1-2 and get knocked out)... you're down to only 3 rounds available for playoffs (before your tiebreaker and 2 finals rounds).

What you could do is, and lemme just name the brackets so what I'm saying is clearer.
Prelims (1-5): Soprano, Alto, Tenor, and Bass, each with 6 teams.
Playoffs (6-7): Soprano 1-2 play Alto 1-2; Tenor 1-2 play Bass 1-2 (carryover a game from prelims; 3-4 and 5-6 all do the same).
Championship (8-9): Girls 1-2 play Boys 1-2 (carryover a game from playoffs).
Advantage final (10-11).
Which leaves one packet available for tiebreaking after prelims or playoffs (you don't need a tiebreaker after the championship phase; ties would either be 3-1-1-1 [clear winner] or 2-2-2-0 [two finals games]).

Essentially, you combine the brackets steadily. If Team X loses its first game to Team Y, they can both win out the rest of the way, and meet up in a final with Team Y holding the advantage.

Is this idea at all clear?

[jonah]

What you could do is, and lemme just name the brackets so what I'm saying is clearer.
Prelims (1-5): Soprano, Alto, Tenor, and Bass, each with 6 teams.
Playoffs (6-7): Soprano 1-2 play Alto 1-2; Tenor 1-2 play Bass 1-2 (carryover a game from prelims; 3-4 and 5-6 all do the same).
Championship (8-9): Girls 1-2 play Boys 1-2 (carryover a game from playoffs).
Advantage final (10-11).
Which leaves one packet available for tiebreaking after prelims or playoffs (you don't need a tiebreaker after the championship phase; ties would either be 3-1-1-1 [clear winner] or 2-2-2-0 [two finals games]).

Essentially, you combine the brackets steadily. If Team X loses its first game to Team Y, they can both win out the rest of the way, and meet up in a final with Team Y holding the advantage.

Is this idea at all clear?

I think so. How do you propose combining brackets for 6-7? The obvious answer would be that the bracket containing the #1 team with the highest PPB plays the bracket containing the #1 team with the lowest PPB, and the middle two play each other, but maybe there's a better way.

What do other people think about this proposal?

My hestitation, and the thing about this format I think is inferior to my original proposal, is that the latter has all the top teams playing all of each other. While it isn't completely critical, I think given our field (and in general, but especially given our field) this would be a really nice feature and I would very much like a way to maximize it.

[jonpin]

You could have pre-determined merges if you were worried about keeping statistics on time or confusing people, but the top teams are probably not going to be confused, so yeah #1 and #4 on PPB (or PPG) go together is probably the best plan.

My hestitation, and the thing about this format I think is inferior to my original proposal, is that the latter has all the top teams playing all of each other.

That's still almost the case. Assuming perfect seeding, the #1 seed plays #8 (in prelims), #4 and #5 (in playoffs), #2 and #3 (in championship). The #3 seed plays #6 (in prelims), #2 and #7 (in playoffs), #1 and #4 (in championship). Each team in the top four plays the other three and two out of the next four. Granted it's not playing all seven of the other top-8, but you're still playing five of them.

[SoLegit12]

Still, no matter how PPG is bad, no one has (or can) refute the argument that PPB distinguishes by enough margins, meaning a secondary indicator should take place. Teams that have 22.43 PPB compared to those 22.41 PPB are virtually similar except for one bonus part

[Mechanical Beasts]

Still, no matter how PPG is bad, no one has (or can) refute the argument that PPB distinguishes by enough margins, meaning a secondary indicator should take place. Teams that have 22.43 PPB compared to those 22.41 PPB are virtually similar except for one bonus part

Look, these sort of arguments are pretty weak. You know what other teams are "virtually similar?" Those that have 241 and 241.1 PPG. That isn't a comment on whether that actually happens at the same tournament very frequently, though!

And as a tiebreaker against PPB, you're proposing something--PPG--that varies with factors other than team skill. So you are literally saying "PPB may correlate more strongly to team knowledge when you must compare across divisions, sure, but since the numbers are sometimes very similar, we'll break ties with a factor that bears less relationship to team knowledge."

[SoLegit12]

Still, no matter how PPG is bad, no one has (or can) refute the argument that PPB distinguishes by enough margins, meaning a secondary indicator should take place. Teams that have 22.43 PPB compared to those 22.41 PPB are virtually similar except for one bonus part

Look, these sort of arguments are pretty weak. You know what other teams are "virtually similar?" Those that have 241 and 241.1 PPG. That isn't a comment on whether that actually happens at the same tournament very frequently, though!

And as a tiebreaker against PPB, you're proposing something--PPG--that varies with factors other than team skill. So you are literally saying "PPB may correlate more strongly to team knowledge when you must compare across divisions, sure, but since the numbers are sometimes very similar, we'll break ties with a factor that bears less relationship to team knowledge."

That still doesn't answer the arguments, sure PPG may occur because of team strength, but I was saying just in case an extremely thin margin in PPB, PPG could be used. Are you going to tell me you could drop a team from potential of 1-6 places to 7-12 because of an extremely slim margin. That team could beat Team B because they could have gotten more powers (not for this tournament), more 30s (just more bonuses; [insert Charlie's bonus chaos argument) and had more tossup knowledge (which are given to every team)? That could happen because of a razor thin margin of PPB; I will get some examples soon.

[Mechanical Beasts]

Still, no matter how PPG is bad, no one has (or can) refute the argument that PPB distinguishes by enough margins, meaning a secondary indicator should take place. Teams that have 22.43 PPB compared to those 22.41 PPB are virtually similar except for one bonus part

Look, these sort of arguments are pretty weak. You know what other teams are "virtually similar?" Those that have 241 and 241.1 PPG. That isn't a comment on whether that actually happens at the same tournament very frequently, though!

And as a tiebreaker against PPB, you're proposing something--PPG--that varies with factors other than team skill. So you are literally saying "PPB may correlate more strongly to team knowledge when you must compare across divisions, sure, but since the numbers are sometimes very similar, we'll break ties with a factor that bears less relationship to team knowledge."

That still doesn't answer the arguments, sure PPG may occur because of team strength, but I was saying just in case an extremely thin margin in PPB, PPG could be used. Are you going to tell me you could drop a team from potential of 1-6 places to 7-12 because of an extremely slim margin. That team could beat Team B because they could have gotten more powers (not for this tournament), more 30s (just more bonuses; [insert Charlie's bonus chaos argument) and had more tossup knowledge (which are given to every team)? That could happen because of a razor thin margin of PPB; I will get some examples soon.

Right, but here's the thing. I'm suggesting that you should use PPB* to make a top-bracket versus second bracket decision, even when . You're suggesting that you should use a statistic that will likely pick the team that, among those two, had the easier prelim bracket. So I guess a thin PPB margin could be a little random; a PPG margin can be random and can be influenced by non-ability factors. Every drama argument about hurting teams applies equally to PPG. (Now, if you're interested in powers, hey--maybe it would be a good idea to pick wildcards based on power rate. I'd want to have statistics to back that up, like the Wynne-Sorice project from... almost a year ago, now? But power rate isn't (very) opponent-dependent, so it may be more fair.)

And let's return to the actual reason we're using PPB. It's to pick a couple of wildcards to advance, along with the first seeds, to the top bracket of the playoffs. Let's say you have six brackets and need to pick two wildcards to make a top bracket of eight. The first three seeds in each of those brackets are:
1, 12, 13
2, 11, 14
3, 10, 15
4, 9, 16
5, 8, 17
6, 7, 18
In almost all cases, you know a lot more about the top several teams in your field than the next group--they've attended more tournaments in the past, they're returning more players from last year's team, or whatever. So those seeds are much more likely to be right. And from the perspective of the PPG a second-place team might get, these brackets are terribly imbalanced. Presumably the second seed in a bracket gets its most competitive games from the first and third seeds, right? So, like, I'm going to assume that mis-seedings from places four on down don't really matter; they're approaching the "playing against chairs" limit for a team's performance. You see the problem, right? If your seeds are all correct, you're basically canonizing the seventh and eighth seeds as the wildcards, which is fine, because your seeding is perfect. But if your seeding is perfect, it's rarely possible for you to do something TOO wrong in a tournament; if your seeding is perfect you don't really have to play games because you can simulate a quintuple round robin in your perfect head and give out plaques. We have to assume mistakes are being made. And if you make a mistake among the second seeds, you could drastically change a team's quality of opponents, thus changing its PPG (but likely not its PPB).

This effect is less prominent for non-second teams, since they're right in the middle of the range where skill varies the most, and in the middle of a region where that variance in skill can strongly affect PPG (since a thirteenth seed is actually probably a lot better at taking tossups away from stronger teams than a thirteenth seed.

* or some better, non-PPG statistic, but I don't think there has been one demonstrated for comparison across brackets

[Irreligion in Bangladesh]

I don't like power rate for a cross-bracket comparison, arguing from an opponent-dependent scenario. If you play against a weak team, you may go more conservative than normal on buzzing to ensure you don't neg yourself out of something they wouldn't have gotten anyway (or at least not until much, much later). Therefore, poor pool construction in the bottom of pools - something that we don't have to worry about as directly with PPB, something that is significantly difficult to get right, and something that really shouldn't make a difference to championship-flight placement - could play a role in how things shake out.

[Stained Diviner]

We won't have powers. Even if we did, power rate is affected by the competition when you have a field as strong as ours is--we have a few teams that can get something like 7 powers per 20 tossups heard on a regular high school set, and that means that people getting beaten will happen often enough to care about.

A separate issue--is there a problem with us using an extra round from another set for a tiebreaker or championship? It looks like both Jonah and Jon's systems leave us a round short if the tiebreakers are necessary. Could we use the last round of HFT if we ended up with a worst case scenario, or is there another packet that would be appropriate?

[Mechanical Beasts]

I don't like power rate for a cross-bracket comparison, arguing from an opponent-dependent scenario. If you play against a weak team, you may go more conservative than normal on buzzing to ensure you don't neg yourself out of something they wouldn't have gotten anyway (or at least not until much, much later). Therefore, poor pool construction in the bottom of pools - something that we don't have to worry about as directly with PPB, something that is significantly difficult to get right, and something that really shouldn't make a difference to championship-flight placement - could play a role in how things shake out.

Yeah, that's a difficulty. (Though they wouldn't play more conservatively given the tiebreakers are known beforehand.) Reinstein's point stands, though, of course.

And as for using HFT--earlier today I told LInda Greene she could distribute the set to coaches. If she hasn't sent that email yet, I guess that would work (unless there were many Auburn kids reading at the Auburn tournament who would therefore have the set already--but I guess they wouldn't have printed rounds they never needed?).

[jonah]

And as for using HFT--earlier today I told LInda Greene she could distribute the set to coaches. If she hasn't sent that email yet, I guess that would work (unless there were many Auburn kids reading at the Auburn tournament who would therefore have the set already--but I guess they wouldn't have printed rounds they never needed?).

We've partially ascertained who has the set and would be relevant to consider for this, and are working on finishing that job. Linda hasn't emailed the set to anyone other than moderators so far, and none of the moderators were current Auburn players. So far it looks like we'd be safe to do that if we wanted to, provided that Stevenson isn't involved in a post-prelims tiebreaker (which is extremely likely, and of course if it happens, we'd do something else). I'm not totally sure that we'd want to do this, Andy, but is it correct that we have your approval and you would not post the set (or publicize discussion thereof) until after the 18th?

[Flutist Wren]

I'm not totally sure that we'd want to do this, Andy, but is it correct that we have your approval and you would not post the set (or publicize discussion thereof) until after the 18th?

Missouri is using the set in May anyway so posting/opening discussion already won't be happening.

EDIT: clarity

[Mechanical Beasts]

I'm not totally sure that we'd want to do this, Andy, but is it correct that we have your approval and you would not post the set (or publicize discussion thereof) until after the 18th?

Missouri is using the set in May anyway so posting/opening discussion already won't be happening.

EDIT: clarity

Oh, that's true, for that singles event.

[Mike Wong]

A separate issue--is there a problem with us using an extra round from another set for a tiebreaker or championship? It looks like both Jonah and Jon's systems leave us a round short if the tiebreakers are necessary. Could we use the last round of HFT if we ended up with a worst case scenario, or is there another packet that would be appropriate?

This is just a thought, but how about the remaining rounds of HSAPQ 15. It doesn't seem like any other mirror played more than 12 rounds, so packets 13 and 14 may be pristine.

[BGSO]

If another packet was going to be needed at some point in the course of the day, would it make sense to arrange the schedules so that the packet affected matches that were as inconsequential as possible (presumably in the first, second, or third round?). That way going into the finals the involved teams don't have to rapidly acclimate to a completely different set of questions.

[Mechanical Beasts]

If another packet was going to be needed at some point in the course of the day, would it make sense to arrange the schedules so that the packet affected matches that were as inconsequential as possible (presumably in the first, second, or third round?). That way going into the finals the involved teams don't have to rapidly acclimate to a completely different set of questions.

So make 1/3 of the three matches that determine whether you're still in championship contention very early on be from a different set? Seems fishy.

Question: repeat checking.

[jonah]

This is just a thought, but how about the remaining rounds of HSAPQ 15. It doesn't seem like any other mirror played more than 12 rounds, so packets 13 and 14 may be pristine.

It got sent to a bunch of teams from Loyburn, so that doesn't work.

If another packet was going to be needed at some point in the course of the day, would it make sense to arrange the schedules so that the packet affected matches that were as inconsequential as possible (presumably in the first, second, or third round?). That way going into the finals the involved teams don't have to rapidly acclimate to a completely different set of questions.

So make 1/3 of the three matches that determine whether you're still in championship contention very early on be from a different set? Seems fishy.

I agree with both of you here. Garb's point would mean we should use it for round 1, when the highest seeds are playing the lowest, but because of the below point, we can't use it for any "regular" round.

Anyway — Kevin Malis has the set, so we can only use a HFT packet for a tiebreaker that doesn't involve Stevenson. A post-prelims situation is the one in which Stevenson is least likely to be involved in a tie, so we'd use it then. (Obviously if Stevenson is in a tie, we'd have to break it on paper or something, but I think it's very unlikely that they will be.)

Question: repeat checking.

Yes. The only packets used at Auburn were 1-11, so assuming no team attending our tournament attended a different HFT-using site (anyone know if that's so?), we can use any of rounds 12, 13, 14, finals 1, finals 2, or the file of extra questions. In all likelihood, between those 5.5 packets there's 20/20 worth of questions that don't substantively overlap anything in GSAC. If we decide to go this route, I'll go through and develop a HFT-derived tiebreaker packet that lacks overlap with GSAC to the greatest extent possible.

[BGSO]

If another packet was going to be needed at some point in the course of the day, would it make sense to arrange the schedules so that the packet affected matches that were as inconsequential as possible (presumably in the first, second, or third round?). That way going into the finals the involved teams don't have to rapidly acclimate to a completely different set of questions.

So make 1/3 of the three matches that determine whether you're still in championship contention very early on be from a different set? Seems fishy.

Question: repeat checking.

I'm just saying, it seems more ideal than deciding the championship game on a packet from a different set.

[jonah]

Carbery and Deveau have it too. We will not be using any HFT questions for tiebreakers (or anything else).

[Aaron Goldfein]

I've always wondered about the logistics of carrying over games from previous rounds. I'll use Jonah's proposed format as an example.

Suppose in one prelim phase pool there are two teams who go 2-1 and two teams who go 1-2. Because the top two teams advance, it is clear and obvious who should go on. However, in the next phase the winner of the game between the 2-1 teams will start 1-0 while the loser will start 0-1. This would be the case even if the loser of that game were declared first in the pool on some sort of statistical measure like PPG. I see this as a fault for the same reason that using head-to-head as a tiebreaker is a fault. That is, it puts extra emphasis on the result of one game.

So, what if instead of simply carrying over the result of the game between the 2-1 teams, you declared that the team who finished in first place entered the next stage 1-0 and the team who finished in second place finished 0-1? It might be more fair to break this tie with the tiebreaker packet (which is available anyway).

A similar issue arises at the conclusion of the parallel-brackets phase. It would probably be more fair to assign the team with the best record as carrying over the record 2-0, the team with the second best record as carrying over 1-1, and the third team as 0-2. Although, if all three teams are tied, then it would probably be best to just assign all teams the record 1-1. And, of course, there would be problems if the top two teams in the pool were tied on record and the third place team was not. I guess you would have to assign the third place team to be 0-2 and the others 2-0 and 1-1 based on statistical measures. I guess you could be really creative and find some system with making them each 1.5-.5, though then you would have to consider finals issues.

[jonpin]

I've always wondered about the logistics of carrying over games from previous rounds. I'll use Jonah's proposed format as an example.

Suppose in one prelim phase pool there are two teams who go 2-1 and two teams who go 1-2. Because the top two teams advance, it is clear and obvious who should go on. However, in the next phase the winner of the game between the 2-1 teams will start 1-0 while the loser will start 0-1. This would be the case even if the loser of that game were declared first in the pool on some sort of statistical measure like PPG. I see this as a fault for the same reason that using head-to-head as a tiebreaker is a fault. That is, it puts extra emphasis on the result of one game.

So, what if instead of simply carrying over the result of the game between the 2-1 teams, you declared that the team who finished in first place entered the next stage 1-0 and the team who finished in second place finished 0-1? It might be more fair to break this tie with the tiebreaker packet (which is available anyway).

A similar issue arises at the conclusion of the parallel-brackets phase. It would probably be more fair to assign the team with the best record as carrying over the record 2-0, the team with the second best record as carrying over 1-1, and the third team as 0-2. Although, if all three teams are tied, then it would probably be best to just assign all teams the record 1-1. And, of course, there would be problems if the top two teams in the pool were tied on record and the third place team was not. I guess you would have to assign the third place team to be 0-2 and the others 2-0 and 1-1 based on statistical measures. I guess you could be really creative and find some system with making them each 1.5-.5, though then you would have to consider finals issues.

Or... you could just look at the parallel playoff bracket as a 6-team round-robin and award each team wins and losses based on whether they beat or lost to each opponent whenever the game may have occurred. Carrying over results of teams that make the same playoff bracket (if you do not plan on re-playing those games) is not at all inconsistent with any QB maxim.

Look at it from the perspective of the two teams in question.
Team A beat Team B, but lost to Team C. Team B beat all others. Team C also lost to Team D. So the records are A 2-1, B 2-1, C 1-2, D 1-2. Now if Team B happens to have more total points, OK, declare them the #1 seed. However, when these two teams get matched up against M and N and X and Y in the playoffs, you can't just give Team A a "loss" against Team B because Team A beat Team B! You have to completely wash away the games against C and D because M,N,X,Y never play them. IMO this is a worse idea than carrying over all games (as NAQT does) as it has the real potential of completely reversing the "outcome" of a game.


To directly counter your statement that such a method overvalues the A-B game, yes and no. Yes, with respect to the playoffs, it counts the A-B game as more important than the A-C, A-D, B-C, and B-D games, because those games are not relevant in the playoff discussion. Counting those games is unfair to teams that wind up in a tough prelim bracket and may take a loss to a team that other playoff teams didn't need to play. No, it doesn't overvalue the A-B game, as it makes it exactly equal to all other games played between teams in the same playoff bracket.

[Aaron Goldfein]

I view each "game" in the parallel bracket phase as the best attempt to decide who is the better of the two teams based on performance. A's head-to-head performance against X, Y, M, and N are the best ways to decide who is better of the two teams. However, team A's performance in a round robin that included B should be a better indicator of which team has performed better and which team is the better of the two than simply the head-to-head game within the round robin.

[Charles Martel]

I view each "game" in the parallel bracket phase as the best attempt to decide who is the better of the two teams based on performance.

The best way to decide who is the better of two teams is to have them play a match.

[jonah]

We're going with what is essentially Jon Pinyan's format. We wrote our own half-packet, so paper tiebreakers should only be necessary in the event of a five-way circle of death after the prelims.

1. Four brackets of six.
2. Break ties (half packet).
3. Three flights of two parallel brackets of four, where each bracket consists of two pairs of teams from the same prelim bracket.
4. Break ties (full packet).
5. Top halves of those parallel brackets play each other, and bottom halves of those parallel brackets play each other.
6. ACF-style finals.

[jonpin]

We're going with what is essentially Jon Pinyan's format. We wrote our own half-packet, so paper tiebreakers should only be necessary in the event of a five-way circle of death after the prelims.

1. Four brackets of six.
2. Break ties (half packet).
3. Three flights of two parallel brackets of four, where each bracket consists of two pairs of teams from the same prelim bracket.
4. Break ties (full packet).
5. Top halves of those parallel brackets play each other, and bottom halves of those parallel brackets play each other.
6. ACF-style finals.

Cool. Just as a note, though, it's entirely possible that the prelims will end with a relevant tie of more than two teams. If there are three teams for one spot, you need more than just a single half-packet.

[jonah]

Yeah. In that case, I think we'd use the full packet then and hope we can get away with the half packet later.

[Aaron Goldfein]

Edit: Ignore this.

[Stained Diviner]

If anybody has a great idea for 22 teams, feel free to share.

[Aaron Goldfein]

I would use a modified version of what you had before.

1. 2 morning brackets of 6, 2 of 5 (brackets of 5 play crossover exhibition matches during their bye)
2. Break ties (half packet)
3. Top two from each bracket form 2 parallel brackets of 4 as before, same with next two, but the rest (6 teams total) form one group
4. As before with top 16 teams

With this system, those who made the top two flights would play 9 games (8 meaningful if in a morning bracket of 5) and those in the bottom flight would play 10 (9 meaningful if in a morning bracket of 5), not including finals and tiebreakers.

[Steeve Ho You Fat]

Or you could just have two brackets of 11 each and then do a A1 v. B2 and B1 v. A2 followed by finals single-elim, and crossovers between the 3rd, 4th, etc. place teams in each bracket. That guarantees 11 games against teams with skill levels from the top to the bottom of the field for everyone.

[jonpin]

Or you could just have two brackets of 11 each and then do a A1 v. B2 and B1 v. A2 followed by finals single-elim, and crossovers between the 3rd, 4th, etc. place teams in each bracket. That guarantees 11 games against teams with skill levels from the top to the bottom of the field for everyone.

1) That's not possible. 11-team RR requires 11 rounds, which leaves only one for a final and NONE for tiebreaking.
2) That's not what he wants. He wants essentially all of the top teams to play each other, with fewer of the "Great Team 585, Outmatched Team 12" results.

[Steeve Ho You Fat]

To be fair, they did write a half packet for tiebreakers, but I forgot 11 team round robin would take 11 packets, not just 10. You're right.