Page 1 of 1

Which tiebreaker: PPG or PPB?

Posted: Wed Jan 09, 2013 8:58 am
by AKKOLADE
Split from discussion of a data entry error in the Centennial tournament thread. --Mgmt.

This is a good time to say that PPB isn't a good tool to use as a tiebreaker for a tournament. Intrabracket PPG is better and fairer.

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 2:10 pm
by Howard
Fred wrote:This is a good time to say that PPB isn't a good tool to use as a tiebreaker for a tournament. Intrabracket PPG is better and fairer.
Is this what you mean?

Within a bracket, PPG is a better tiebreaker because it better represents actual average performance in a game.

Between brackets, PPB is a better tiebreaker because it removes differences resulting from playing different opponents.

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 7:46 pm
by Pushkin's Beard
Fred wrote:This is a good time to say that PPB isn't a good tool to use as a tiebreaker for a tournament. Intrabracket PPG is better and fairer.
But PPG isn't really a fair tiebreaker because it could just come down to who beat up on the weakest team more.

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 7:52 pm
by AKKOLADE
Within the same bracket, PPG is better because it gives a reasonable measurement of team skill levels. "Beating up on weak teams" isn't a good argument; it could just as well be "scoring against good teams" or scoring in general. PPB is not as good because it can depend on statistical oddities and uses a smaller sample size.

Between different brackets, PPG and PPB are both flawed. PPG involves playing different opponents, making comparisons difficult at best, while PPB has the same issues of "a handful of rounds from a single tournament." If I had to use something, I probably would use PPG.

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 8:26 pm
by Mewto55555
PPB is also far far easier to screw up in calculation, which would be bad if it actually happened (and I think has? anytime you have to go to SQBS and can't easily check by hand you're looking for problems to come your way).

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 8:41 pm
by dtaylor4
Mewto55555 wrote:PPB is also far far easier to screw up in calculation, which would be bad if it actually happened (and I think has? anytime you have to go to SQBS and can't easily check by hand you're looking for problems to come your way).
At New Trier's 2011 tournament (where all protests were resolved, regardless of game outcome), PPB was used to decide top bracket wild cards. Carbondale had 25.09 PPB, DCC had 25.08. To add, New Trier and Belvidere North A had higher PPBs, but worse records.

Re: Centennial High School Tournament (12/2/12)

Posted: Fri Jan 11, 2013 9:01 pm
by AKKOLADE
There was a Texas tourney, I think last year, that had Cistercian eliminated due to a math error.

Fred's syntax corrected; subsequent posts FZed --JTH

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 12, 2013 1:27 am
by jonpin
I see the arguments for using PPB in picking "wild cards" where not all teams at a certain level will advance. However, using PPB to rank tied teams within a round-robin is, in my opinion, wrong and should stop. Quiz bowl is a game of tossups and bonuses, and if your ranking criteria uses only bonuses, it is ignoring more than half of the game data available. There is no "strength of schedule" argument to be used against PPG / total points within a round-robin, and there are several arguments against using PPB:

1. Unlike PPG / total points, it is not reasonable to calculate by hand "on the fly". At tournaments I run, it is my policy to spend the last round looking at places where ties could develop and finding the total points of those teams entering the final round. That way, when results come in, I can add in the points from the last round and instantly know the ranking of those teams on total points (which is obviously equivalent to the ranking on PPG). That way I can start typing into the playoff schedules while the scoresheets are entered into SQBS (which then confirms my numbers). PPB calculations are more difficult arithmetic, and cannot be done in a matter of seconds. Given the long lengths of time it takes tournaments to rebracket, the quicker you have the rankings settled (or you have the seeds for tiebreakers ready), the better.

2. Because PPG is a very simple calculation, it is harder to get wrong in SQBS, and it is easier to fix. Unless you mistype (or misread) the final score of a game, PPG will not be wrong. There are many more data points that can lead to getting PPB wrong. Worse, errors in PPB almost always are compounded. If a team actually scored 130 points on 10 bonuses, and some player Mark went 3/0, but you gave Mark a line of 2/0, SQBS calculates 140 points on 9 bonuses. Thus you've increased the numerator AND decreased the denominator. Likewise if you gave Mark a line of 0/3, SQBS calculates 175 points on 7 bonuses, an even bigger error in both directions. Even still, it's very unlikely that looking at the stats you will be able to instantly tell this is wrong. From a practical viewpoint, this is my biggest argument against PPB: it's easy to get wrong.

3. The "beating up on weak teams" argument is terrible and counterproductive. PPG always rewards scoring points. Obviously. PPB does not. Let's say you're entering the last round of prelims. You're playing a weak team and expect to win, and Central HS is playing a strong team and expected to lose; this will cause a tie in the standings between you and Central for the last playoff spot. If PPG is the tiebreaker, then your goal is simple: score enough points to make up any deficit to Central, so that you'll have scored more over the course of the round-robin. It is always to your benefit to get a tossup correct, because it is points and access to a bonus. If PPB is the tiebreaker, and you see from the standings that your average is currently 23, ahead of Central's 21, once you've secured the victory, it can be to your benefit to not answer tossups, as you risk jeopardizing your high PPB. If you get the last three tossups, but wind up with 0s or 10s on the bonuses, that can drag your average behind Central's. You'd have been better off putting the buzzer down and not answering the tossup at all. Admittedly, this is a contrived example, but the fact that there exists a scenario with PPB as a tiebreaker, where not providing a correct answer can be a strategic advantage should be a disqualifier for this tiebreaker where better alternatives exist.

To cite the New Trier example, here are the stats. If Carbondale had gotten one more tossup correct and 20 on the bonus, they would be eliminated. I find it not good to have a situation where the getting that tossup correct is ignored or in fact potentially penalized in determining who advances.

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 12, 2013 2:48 am
by Realdeo
Pardon, but what's PPG and PPB? Is it point per group and point per bracket?

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 12, 2013 2:56 am
by dtaylor4
Realdeo wrote:Pardon, but what's PPG and PPB? Is it point per group and point per bracket?
Points per game and points per (controlled) bonus.

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 12, 2013 10:54 am
by Marble-faced Bristle Tyrant
I was just trying to explain to someone the use of ppb tiebreakers, and then this thread shows up. Welp.

Re: Which tiebreaker: PPG or PPB?

Posted: Sun Jan 13, 2013 10:53 am
by Dewey
I have always thought PPB and then # of Powers should be the tiebreaker. PPB because it is often considered to be the standard in determining team ranking, etc. To me it just seems more optimal because it is not strength of competition based. If the PPB is close though then # of Powers should be used because, at least in my mind, when you are playing with a relatively average field you can expect that powering a question would result in you keeping said question in most games. If the tournament doesn't have powers then that wouldn't work as well.

Re: Which tiebreaker: PPG or PPB?

Posted: Sun Jan 13, 2013 11:35 am
by AKKOLADE
Dewey wrote:I have always thought PPB and then # of Powers should be the tiebreaker. PPB because it is often considered to be the standard in determining team ranking, etc. To me it just seems more optimal because it is not strength of competition based. If the PPB is close though then # of Powers should be used because, at least in my mind, when you are playing with a relatively average field you can expect that powering a question would result in you keeping said question in most games. If the tournament doesn't have powers then that wouldn't work as well.
Ignoring all the arguments laid out above, how close is "close" for PPB?

The reason PPB is "the" standard in determining team ranking is because it allows you to compare the performances of teams at different tournaments, on different sets, across the entirety of the season. Those strengths for it are less so when you're talking about a single tournament.

Powers are just as easily affected, if not more so, by strength of opponent as PPG is. It also relies on a smaller sample size of questions.

Re: Which tiebreaker: PPG or PPB?

Posted: Sun Jan 13, 2013 5:59 pm
by Song Oku
In cases where PPB is the tiebreaker, is head-to-head record accounted for at all? For example, if Team A and Team B were both 5-4, with Team A with higher PPB but Team B beat Team A, then who would be placed higher?

Re: Which tiebreaker: PPG or PPB?

Posted: Sun Jan 13, 2013 6:10 pm
by Important Bird Area
The head-to-head result is already accounted for in the team records; using it as an additional tiebreaker effectively double-counts a single game (and has the potential to reward teams that sustain upset losses to weak opponents). See discussion and statistical analysis here.

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 26, 2013 8:46 pm
by Capitoline
Dewey wrote:I have always thought PPB and then # of Powers should be the tiebreaker. PPB because it is often considered to be the standard in determining team ranking, etc. To me it just seems more optimal because it is not strength of competition based. If the PPB is close though then # of Powers should be used because, at least in my mind, when you are playing with a relatively average field you can expect that powering a question would result in you keeping said question in most games. If the tournament doesn't have powers then that wouldn't work as well.
I don't think total number of powers is feasible for widespread use because of the difference in difficulty between sets. A team might power 5 times per game on an A set while a better team powers 3 times on an IS set. In addition, some teams simply play more conservatively than others. This strategic choice shouldn't hinder a team's ranking/placement.

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 26, 2013 8:54 pm
by AKKOLADE
I think we're mostly talking about tiebreakers for seeding/reseeding an individual tournament.

Re: Which tiebreaker: PPG or PPB?

Posted: Sat Jan 26, 2013 9:00 pm
by Important Bird Area
(maybe time for a thread split?)

And when we are talking about comparing performance on different sets:

1) HSNCT seeding corrects for the relative difficulty of the different sets played by different teams.

and

2) Insofar as powering questions is a strategic choice, that choice should absolutely be taken into account in seeding and ranking teams.

Re: Which tiebreaker: PPG or PPB?

Posted: Sun Jan 27, 2013 9:17 am
by Capitoline
bt_green_warbler wrote:
2) Insofar as powering questions is a strategic choice, that choice should absolutely be taken into account in seeding and ranking teams.
I think powers (if they have to be used at all in seeding) should only be taken into account in the form of power/neg ratio. Otherwise, a team with average knowledge feeling aggressive on the buzzers might end up seeded higher than a good team playing at a normal pace, as a result of getting both more powers and more negs. Powering questions could either be a result of skill or a result of luck, with no way to differentiate between the two by stats, unless the ratio is used.