How to improve the Pairwise

[Shirtless Guy]

Re: How to improve the Pairwise

Listening to today's USCHO Live recording. http://www.uscho.com/uscho-live/2016/04/07/uscho-live-thursday-at-the-2016-ncaa-frozen-four/

HEA commish Joe Bertagna did a fair amount of bloviating about the selection process, saying that there should basically be more wiggle room to go against the coaches' wishes to have transparency in the selection/seeding process. Basically in favor of a more basketball-like closed-door, smoky boardroom sort of shenanigans... specifically taking into account how a team is playing in the last X number of games to fudge stuff around.

Correct me if I'm wrong, but is there really anyone on the boards here who thinks that way?

Whatever arguments we might have over the numerical method chosen, I feel like we can all agree that we like the transparency of a purely numerical method.

I would be ok with more fudgability to change swap 8 and 9 or 12 and 13 vs having hard bands (should it be extremely helpful to attendance, avoiding interconference match, etc) but beyond that, I think we all agree that a transparent system is important...

 

[Patman]

Listening to today's USCHO Live recording. http://www.uscho.com/uscho-live/2016/04/07/uscho-live-thursday-at-the-2016-ncaa-frozen-four/

HEA commish Joe Bertagna did a fair amount of bloviating about the selection process, saying that there should basically be more wiggle room to go against the coaches' wishes to have transparency in the selection/seeding process. Basically in favor of a more basketball-like closed-door, smoky boardroom sort of shenanigans... specifically taking into account how a team is playing in the last X number of games to fudge stuff around.

Correct me if I'm wrong, but is there really anyone on the boards here who thinks that way?

Whatever arguments we might have over the numerical method chosen, I feel like we can all agree that we like the transparency of a purely numerical method.

Transparency is a big deal. But the method has its difficulties as well... Applying the pairwise to basketball, for instance, would produce results that would give people a lot of pause.

In the end, I favor the ultimate transparency of a formulaic driven result. It does remove the drama and I think sports people don't like it because it makes the world too clean and it doesn't work with their own preconceptions.

 

[The Sicatoka]

Re: How to improve the Pairwise

Q: How to improve the Pairwise?

A: Don't. Scrap it. Use KRACH.

 

[ExileOnDaytonStreet]

Re: How to improve the Pairwise

I don't see why KRACH means you have to replace the PWR. Just use KRACH instead of RPI.

The only complaints that I really have with the current method is that RPI is a more finicky metric.

 

[tape]

Re: How to improve the Pairwise

I don't see why KRACH means you have to replace the PWR. Just use KRACH instead of RPI.

The only complaints that I really have with the current method is that RPI is a more finicky metric.

Heck, just add KRACH to the existing criteria. If you do that, you're going to have at least 3 criteria for every team comparison, instead of the 2 criteria for 70% of the comparisons that we have now which results in the RPI being vastly overweighted.

 

[manurespreader]

Re: How to improve the Pairwise

I agree that Kratch could be added but otherwise I want a transparent system.

I still believe in making the QWB a geometric progression instead of an arithmetic one. Also, Brad Schlossmann said yesterday that at the end of the season, UND was not really playing as hard as it might have otherwise. He pointed to the consolation game of the NCHC tournament, and said both teams just went thru the motions. I agree with that, and it's not just UND that does it. So I would find a way to lesson the points given to a team that beats another team who has clinched. It's not the same quality of win and it has an effect out of proportion to what is deserved.

 

[joecct]

Re: How to improve the Pairwise

I think winning your conference removes all doubt.

At-Large teams hurt the intensity of the conference tournaments. After all, if you're 4th in the PWR, a loss won't knock you out. Remove the at-large and even #1 has to work to get to the NCAA tournament.

 

[Craig P.]

Re: How to improve the Pairwise

For example...the result of every game directly impacts the weight of Opp's W% and OppOpp W%. Should UMD's Opp W% be determined on whether or not they win a game? Or does it make sense that the 2nd and 3rd components of RPI should not be weighted based on the outcome of the game?

Unless this has changed recently, your own games against an opponent are excluded from the OppW% column. It's one of the things that has always made RPI deceptively tricky to calculate.

My issues with it are as follows:
I think the OppOpp is weighted too high and I'd find some way to lower it slightly.

Just because the multiplier is large, doesn't mean that it has a large effect on the RPI. There is relatively little variation in OppOpp from one team to another, so even with an apparently large multiplier, it still may not have that much influence.

 

[Shirtless Guy]

Re: How to improve the Pairwise

Unless this has changed recently, your own games against an opponent are excluded from the OppW% column. It's one of the things that has always made RPI deceptively tricky to calculate.



Just because the multiplier is large, doesn't mean that it has a large effect on the RPI. There is relatively little variation in OppOpp from one team to another, so even with an apparently large multiplier, it still may not have that much influence.

You're not understanding what I'm saying...if a team wins a home game, that game gets a game weight of 0.8, if they lose its 1.2. That changes the weight of the OppW% and OppOppW%. While the game result itself doesn't change the OppW%, it does change the weight of the values, which is confusing.

 

[Finnedog]

Re: How to improve the Pairwise

I think winning your conference removes all doubt.

At-Large teams hurt the intensity of the conference tournaments. After all, if you're 4th in the PWR, a loss won't knock you out. Remove the at-large and even #1 has to work to get to the NCAA tournament.

To me winning your conference's regular season would give a better indication of how you would fair in the NCAA tourney; but if you take away the possibility that a bad team might get hot and win their conference tourney a majority of the teams would have no chance to make the tournament after half the season is over. The conference tourney is a new season for them.
I do agree with a high at large seed not needing to play hard at conference tourneys, see Denver and North Dakota.

 

[Wisko McBadgerton]

Re: How to improve the Pairwise

I think winning your conference removes all doubt.

At-Large teams hurt the intensity of the conference tournaments. After all, if you're 4th in the PWR, a loss won't knock you out. Remove the at-large and even #1 has to work to get to the NCAA tournament.

Then why not just eliminate conference tourneys altogether (no one is showing up to watch half of them anyway) and put everyone in. Bye the top 4, best two out of three in the first round at the higher seed, or even the first two rounds. ( would still only give you a max of 10 postseason games possible for the champs.) That would greatly reduce the pressure on the PWR getting it perfect because even if you're screwed by the system, you're not out, and would just generally make for a great tournament with zero meaningless post season games. It would bring in the real Cinderella possibility and also largely account for teams that are playing their best down the stretch. You could also probably restore some traditional rivalries lost in realignment by breaking into 4 regions. Who wouldn't love that?

 

[joecct]

Then why not just eliminate conference tourneys altogether (no one is showing up to watch half of them anyway) and put everyone in. Bye the top 4, best two out of three in the first round at the higher seed, or even the first two rounds. ( would still only give you a max of 10 postseason games possible for the champs.) That would greatly reduce the pressure on the PWR getting it perfect because even if you're screwed by the system, you're not out, and would just generally make for a great tournament with zero meaningless post season games. It would bring in the real Cinderella possibility and also largely account for teams that are playing their best down the stretch. You could also probably restore some traditional rivalries lost in realignment by breaking into 4 regions. Who wouldn't love that?

I actually am old enough to remember when the NCAA hoops tournament was 1 team per conference. The ACC was the only conference with a tournament that determined the NCAA team. It was intense. It was sold out for 3 nights. It was epic.

I also remember when the NCAA was 4 teams and the ECAC 1st and 2nd place tournament finishers (with the notable exception of BU screwing over Clarkson in 1971) were the ECAC representatives. If #1 lost before the finals, they were done. It made the tournament intense and the Garden was rocking.

Why not return to that feeling?

 

[billmich88888]

Re: How to improve the Pairwise

I actually am old enough to remember when the NCAA hoops tournament was 1 team per conference. The ACC was the only conference with a tournament that determined the NCAA team. It was intense. It was sold out for 3 nights. It was epic.

I also remember when the NCAA was 4 teams and the ECAC 1st and 2nd place tournament finishers (with the notable exception of BU screwing over Clarkson in 1971) were the ECAC representatives. If #1 lost before the finals, they were done. It made the tournament intense and the Garden was rocking.

Why not return to that feeling?

Because back then you had basically only 2 conferences, so these games you speak of were defacto national tournament games,

with 6 conferences today, cant get that genie back in the bottle

 

[Craig P.]

Re: How to improve the Pairwise

You're not understanding what I'm saying...if a team wins a home game, that game gets a game weight of 0.8, if they lose its 1.2. That changes the weight of the OppW% and OppOppW%. While the game result itself doesn't change the OppW%, it does change the weight of the values, which is confusing.

Understood, thanks.

 

[Patman]

Understood, thanks.

Hmm... I wonder if that's what was wrong with my RPI code...

Why do you make things so needlessly difficult, NCAA?

 

[Jasma]

Re: How to improve the Pairwise

Changing the NCAA tourney format might liven it up.
One scenario:
A 16 team NCAA Tourney, as it is now.
The finalists of the 6 leagues make the tourney - 12 teams
There are 4 at-large selections made by the committee
The committee then seeds the teams and places them in the four regionals
The 4 at-large berths help protect strong teams that get upset in early-round conference tournament games.
This format does add one more Atlantic Hockey team to the NCAAs, but it also livens up the conference tourneys. Even the finals could be important for seeding.
Oh, and move the regionals to smaller venues. this might be easier to do in the Northeast, but I'm not familiar with midwest and western cities with smaller arenas.
In the northeast you have cities like Bridgeport, Trenton, Syracuse, Glens Falls, Rochester, Manchester that seat fewer than places like Albany, Providence and Worcester. 4,000 to 6,000 fannies in a building that seat 8-10,000 is a lot better than what we have seen in the larger buildings and cities.
Flaws in this I'm sure, but something needs to change. There is not enough excitement at most of the conference and regional tournaments the way it is set-up now.

 

[ExileOnDaytonStreet]

Re: How to improve the Pairwise

Changing the NCAA tourney format might liven it up.
One scenario:
A 16 team NCAA Tourney, as it is now.
The finalists of the 6 leagues make the tourney - 12 teams
There are 4 at-large selections made by the committee
The committee then seeds the teams and places them in the four regionals
The 4 at-large berths help protect strong teams that get upset in early-round conference tournament games.
This format does add one more Atlantic Hockey team to the NCAAs, but it also livens up the conference tourneys. Even the finals could be important for seeding.
Oh, and move the regionals to smaller venues. this might be easier to do in the Northeast, but I'm not familiar with midwest and western cities with smaller arenas.
In the northeast you have cities like Bridgeport, Trenton, Syracuse, Glens Falls, Rochester, Manchester that seat fewer than places like Albany, Providence and Worcester. 4,000 to 6,000 fannies in a building that seat 8-10,000 is a lot better than what we have seen in the larger buildings and cities.
Flaws in this I'm sure, but something needs to change. There is not enough excitement at most of the conference and regional tournaments the way it is set-up now.

I stopped reading this pretty quickly.

TWO Big Ten teams?

 

[Superfly100]

Changing the NCAA tourney format might liven it up.
One scenario:
A 16 team NCAA Tourney, as it is now.
The finalists of the 6 leagues make the tourney - 12 teams
There are 4 at-large selections made by the committee
The committee then seeds the teams and places them in the four regionals
The 4 at-large berths help protect strong teams that get upset in early-round conference tournament games.
This format does add one more Atlantic Hockey team to the NCAAs, but it also livens up the conference tourneys. Even the finals could be important for seeding.
Oh, and move the regionals to smaller venues. this might be easier to do in the Northeast, but I'm not familiar with midwest and western cities with smaller arenas.
In the northeast you have cities like Bridgeport, Trenton, Syracuse, Glens Falls, Rochester, Manchester that seat fewer than places like Albany, Providence and Worcester. 4,000 to 6,000 fannies in a building that seat 8-10,000 is a lot better than what we have seen in the larger buildings and cities.
Flaws in this I'm sure, but something needs to change. There is not enough excitement at most of the conference and regional tournaments the way it is set-up now.

Building on the top two from each conf tournament, but with a twist...

Drop from four regionals to two with set conferences in each regional. NCHC, WCHA, and BI6(7) in the West and HE, ECAC, and AHA in the East. Two At-Large selections for each regional. Tournaments held at a central location for each Regional.

It would take the PWR out of the picture, build on inter-conference rivalries, and stoke the East vs West rivalry.

 

[Gr8Sk8M8]

Re: How to improve the Pairwise

At the risk of being redundant, a conference winner needs to be seeded higher than any other team within its conference outside of a Top 4 finish.

This is why Pairwise is an absolutely moronic system...there are several teams in Major League Baseball that have won Division titles but who absolutely ---ed the bed against last place teams. Using a Pairwise system, in years past, the Red Sox, Yankees and Cardinals would have not gotten a playoff berth in a year that they won the World Series.

NU beat Harvard and NU won the conference. NU had to travel ~900 miles and play the top seed in the bracket; Harvard, who won nothing, had to travel 40 and was seeded 3rd. Yale didn't get out of their quarterfinals and got rewarded with Albany and a 3rd seed.

Those two examples show the absolute absurdity of Pairwise.

 

[FlagDUDE08]

Re: How to improve the Pairwise

You replace this system with another system. The next year, we see another one of these stupid threads about ways to improve that system.

Suck it up. You know the formula. If you're really worried about it, how about focusing more on winning your league, which is the sure fire way of getting into the national tournament.