The 2014 Pairwise, Bracketology and History Thread`

[ExileOnDaytonStreet]

Re: The 2014 Pairwise, Bracketology and History Thread`

The NCHC is similar. 6 of the 8 teams are above .500 OOC, going a combined 32-14-12. The 3-16-3 record of CC and UNO is what hurts.

What's the NCHC's conference schedule? It's not a full quadruple round-robin like the BTHC, is it (I'm only seeing CC on NoDak's schedule twice)? Because that's one thing that will hurt the B1Gs is the fact that everyone has to play Penn State and Michigan State four times, whereas some NCHC teams only have to play 6 games against the worst in their conference.

 

[ExileOnDaytonStreet]

Re: The 2014 Pairwise, Bracketology and History Thread`

If I followed your logic correctly, applying to the recent tournament outcomes, then the ECAC should end up with 6 teams in the tournament and all four slots in the Frozen 4..

My logic was more of a "comparison against the mean". That is, I looked at the average of what these programs have done over the last decade and used that as a benchmark for what each conference would expect as an "average" result. Showing better or worse results than said benchmark would be indicative of having an above average or below average season for the conference. Other than Quinnipaic, the ECAC has remained unchanged for quite some time, making it one of the more fair conferences to apply that logic to.

The logic you're presenting is more of the the "if present trends continue..." variety. And aside from saying that those are two wildly different logical processes/sets of assumptions, I'll just stand on a soapbox and say that the "trends" type of logic requires a hell of a lot more statistical evidence to back up than the assumption that everyone regresses to a mean eventually.

All that said:

My assumptions are going to be not-totally accurate in that the new set up out West is going to lead to teams that are used to more success in the WCHA and CCHA potentially end up towards the bottom of the B1G and NCHC while programs that were weaker in their conferences before realignment could end up at the top of the nWCHA. There's a certain amount of movement to be expected amongst those conferences, since they are clearly in wildly different circumstances than they were in as members of the old WCHA and the CCHA (although their overall strengths/weaknesses as programs might lead one to expect reasonably similar OOC results).

On the other hand, HE, ECAC and AHA have all stayed relatively stable in recent years, and any movement that those conferences make will be almost entirely due to (a) the merits of Notre Dame in Hockey East and (b) any general improvements/regressions in their OOC performance. Obviously, we've seen ECAC improve overall as a conference in the past few seasons, and this year there is absolutely no reason to suspect otherwise. I don't see how it's unreasonable to say that this year is "above average" for them. Perhaps I could have added that as an improving conference, they have the potential to set higher expectations (bump up their "average" result), but we only have a few seasons of improvement to work with here, and long-term averages are not erased so easily.​

 

[ExileOnDaytonStreet]

Re: The 2014 Pairwise, Bracketology and History Thread`

Copied from GPL and obviously doesn't incorporate context, but found it interesting otherwise:

The irony is, that the changes made in the pairwise pretty much make the 'pairwise' obsolete. The pairwise now is pretty much just the RPI, since there is no more TUC comparison. It takes two other comparisons to override the RPI. Without the TUC, now there are only two other comparisons, common opponent and head-to-head. So if two teams don't play each other, the lower RPI team can't overtake the higher RPI team, even if they have a better record against common opponents. Looking at the USCHO pairwise, the pairwise order is exactly the RPI order for 1-24. Denver loses a couple comparisons due to some HTH losses combined with COp in one case. The point is, all of the nice interconference games we are finally getting are becoming less meaningful now. Of course, they all count for RPI.

It seems to me that the first tiebreaker ought to be H2H, if applicable. Isn't that the basic idea of the pairwise, each team considered individually against every other team? Moreover, I'm annoyed that they de-emphasized H2H (don't know quite when - this year?). Used to be if I beat you 3-0 during the season I received 3 comparison points rather than the measly one I would get now. If the point is each pair of teams compared in a vacuum, shouldn't a season sweep be strongly rewarded? With the schedule scrambling that occurred in HE this year, some teams could meet six or seven times. Wouldn't 6-0 or 5-2 in favor of one team be more indicative than the combination of RPI and common opponents for that particular pair?

How come the guys in charge of this stuff aren't as bright as us board denizens?

I do think it'd be fairly interesting to see how the PWR for this season would compare under any of the following tweaks:

- Bring back the TUC comparison, and set the TUC requirement to an RPI above .500.
- Bring back the TUC and do the top 25 teams.
- Change the H2H comparison to be one full comparison point per win.
- Switch the tiebreaker to H2H.

Or, and perhaps this is simpler, are there any sites that list RPI and PWR at the end of the season before the NCAAs? It'd be interesting to see how closely the PWR followed the RPI rankings in other years before the removal of TUC.

 

[Numbers]

Re: The 2014 Pairwise, Bracketology and History Thread`

I do think it'd be fairly interesting to see how the PWR for this season would compare under any of the following tweaks:

- Bring back the TUC comparison, and set the TUC requirement to an RPI above .500.
- Bring back the TUC and do the top 25 teams.
- Change the H2H comparison to be one full comparison point per win.
- Switch the tiebreaker to H2H.


Or, and perhaps this is simpler, are there any sites that list RPI and PWR at the end of the season before the NCAAs? It'd be interesting to see how closely the PWR followed the RPI rankings in other years before the removal of TUC.

Quick answer is that Whelan's site had it, but I can't find archives to go further back than last year. Last year anyway, it was like this:
7 & 8 were flip flopped
9th in RPI was 11th in PWR
12th in RPI was 14th in PWR
Robert Morris, at 17th in PWR, actually jumped about 10 spots.

Not sure what to conclude, but that's the results among the top 20 RPI teams.

 

[SJHovey]

What's the NCHC's conference schedule? It's not a full quadruple round-robin like the BTHC, is it (I'm only seeing CC on NoDak's schedule twice)? Because that's one thing that will hurt the B1Gs is the fact that everyone has to play Penn State and Michigan State four times, whereas some NCHC teams only have to play 6 games against the worst in their conference.

24 conference games.

 

[QUAlum2004]

I came up with a slightly different bracket than Priceless, but there's surely a tough decision for the committee this week. Punishing Minnesota by playing the #13 seed is a tough call. Beyond that, moving QU out of Bridgeport and BC out of Worcester really hurts attendance.

East - Bridgeport (Yale)
2. Boston College
6. Providence
12. Clarkson
14. Yale

Northeast - Worcester
3. Quinnipiac
7. Northeastern
10. Wisconsin
16. Mercyhurst

Midwest - Cincinnati (Miami)
4. Union
5. Ferris State
11. Notre Dame
15. Michigan

West - St. Paul (Minnesota)
1. Minnesota
8. St. Cloud State
9. UMass-Lowell
13. Cornell

Quinnipiac would still send a lot of people to Worcester.

 

[burgie12]

Re: The 2014 Pairwise, Bracketology and History Thread`

I do think it'd be fairly interesting to see how the PWR for this season would compare under any of the following tweaks:

- Bring back the TUC comparison, and set the TUC requirement to an RPI above .500.
- Bring back the TUC and do the top 25 teams.
- Change the H2H comparison to be one full comparison point per win.
- Switch the tiebreaker to H2H.

Or, and perhaps this is simpler, are there any sites that list RPI and PWR at the end of the season before the NCAAs? It'd be interesting to see how closely the PWR followed the RPI rankings in other years before the removal of TUC.

Head-to-head is still one comparison point per win. So, if Team A beats Team B three times and ties the fourth game (assuming they play four games), they can't lose the pairwise comparison between the two teams. (Unless I'm misunderstanding your third bullet...)

 

[ExileOnDaytonStreet]

Head-to-head is still one comparison point per win.

Ah, and so it is.

Nevermind about the H2H stuff.

Although it would still be interesting to see H2H as the tiebreaker instead of RPI.

 

[Steve_MN]

Re: The 2014 Pairwise, Bracketology and History Thread`

Ah, and so it is.

Nevermind about the H2H stuff.

Although it would still be interesting to see H2H as the tiebreaker instead of RPI.

I'd definitely like to see H2H being the first tiebreaker within comparisons, then RPI for season splits and teams that didn't face each other.

Interesting write-up on the conferences compared to the last 10 years.

Now cut that out, you're making me agree with a Wisconsin fan

 

[burgie12]

Re: The 2014 Pairwise, Bracketology and History Thread`

I'd definitely like to see H2H being the first tiebreaker within comparisons, then RPI for season splits and teams that didn't face each other.

Interesting write-up on the conferences compared to the last 10 years.

Now cut that out, you're making me agree with a Wisconsin fan

I'd also prefer it if the Pairwise Comparison itself were the tiebreaker when two teams end up with the same number of comparison wins, but let's not get too carried away fixing everything that the committee does wrong.

 

[JimDahl]

Re: The 2014 Pairwise, Bracketology and History Thread`

Got the simulator up and running, some general observations from my first forecast through the end of the regular season:

Pairwise Rankings and the NCAA hockey tournament outlook

Alas, other higher priorities (including implementing the new algorithms for PWR) mean I'm still doing a break between the regular season and conference tournaments.

 

[TonyTheTiger20]

Re: The 2014 Pairwise, Bracketology and History Thread`

Got the simulator up and running, some general observations from my first forecast through the end of the regular season:

Pairwise Rankings and the NCAA hockey tournament outlook

That's good stuff.

 

[Numbers]

Re: The 2014 Pairwise, Bracketology and History Thread`

Sat Eve Jan 25. 11:11 CT (Alaska games late, I don't have a simulator....)

#1 seeds: Minn, BC, Quinnipiac, Union
#2: Ferris, St Cloud, Providence, Northeastern
#3: Lowell, Michigan, Cornell, Notre Dame
#4: Wisconsin, UMD, Colgate, AHA Champion

St Paul: Minn v AHA; Northeastern v Michigan
Worcester: BC v Colgate; St Cloud v Lowell
Bridgeport: Quinnipiac v UMD; Providence v Cornell
Cincinnati: Union v Wisconsin; Ferris State v Notre Dame

Explanation of bracket changes:
1) For the 4th seeds: I put UMD in Worcester and Colgate in Bridgeport because Colgate will get a few more people to Bridgeport (Sorry I don't know the Eastern fan bases well, but this seems to make sense to me). NOW EDITED: QU and Colgate can't play each other, so BC v Colgate, and QU v UMD.
2) For the 2nd and 3rd seeds: Ferris has to go to Cincinnati. Put either Michigan or Notre Dame there for attendance. Notre Dame makes it the right bracket, but I wouldn't quibble either way. After that, Lowell has to play St Cloud because of NE and Prov being HE teams. I put Northeastern (8th overall) in St Paul because of bracket integrity. It makes sense to put Michigan there, too, to get as many eastern #3 seeds in the eastern regionals as possible. So, after NE v Mich goes to St Paul, I put St Cloud v Lowell in Worcester and Providence v Colgate in Bridgeport.

Someone help me out if there is a better way for attendance for the eastern schools:

I mean,
Is N'Eastern a better draw than Providence for Bridgeport?
Is, for example, the best arrangement to put N'Eastern v Colgate in Worcester, and St Cloud v Lowell in Bridgeport?

I would be happy if someone educated me.

Thanks.

 

[Patman]

Re: The 2014 Pairwise, Bracketology and History Thread`

Sat Eve Jan 25. 11:11 CT (Alaska games late, I don't have a simulator....)

#1 seeds: Minn, BC, Quinnipiac, Union
#2: Ferris, St Cloud, Providence, Northeastern
#3: Lowell, Michigan, Cornell, Notre Dame
#4: Wisconsin, UMD, Colgate, AHA Champion

St Paul: Minn v AHA; Northeastern v Michigan
Worcester: BC v UMD; St Cloud v Lowell
Bridgeport: Quinnipiac v Colgate; Providence v Cornell
Cincinnati: Union v Wisconsin; Ferris State v Notre Dame

Explanation of bracket changes:
1) For the 4th seeds: I put UMD in Worcester and Colgate in Bridgeport because Colgate will get a few more people to Bridgeport (Sorry I don't know the Eastern fan bases well, but this seems to make sense to me)
2) For the 2nd and 3rd seeds: Ferris has to go to Cincinnati. Put either Michigan or Notre Dame there for attendance. Notre Dame makes it the right bracket, but I wouldn't quibble either way. After that, Lowell has to play St Cloud because of NE and Prov being HE teams. I put Northeastern (8th overall) in St Paul because of bracket integrity. It makes sense to put Michigan there, too, to get as many eastern #3 seeds in the eastern regionals as possible. So, after NE v Mich goes to St Paul, I put St Cloud v Lowell in Worcester and Providence v Colgate in Bridgeport.

Someone help me out if there is a better way for attendance for the eastern schools:

I mean,
Is N'Eastern a better draw than Providence for Bridgeport?
Is, for example, the best arrangement to put N'Eastern v Colgate in Worcester, and St Cloud v Lowell in Bridgeport?

I would be happy if someone educated me.

Thanks.

I'd say PC would draw much better in Bridgeport than NU. Its been longer away from the tournament for PC and a much shorter driving distance.

 

[ScottK]

Re: The 2014 Pairwise, Bracketology and History Thread`

I'd say PC would draw much better in Bridgeport than NU. Its been longer away from the tournament for PC and a much shorter driving distance.

It's 45 minutes to drive from Providence to Worcester and 2 hours from Providence to Bridgeport.

 

[Numbers]

Re: The 2014 Pairwise, Bracketology and History Thread`

It's 45 minutes to drive from Providence to Worcester and 2 hours from Providence to Bridgeport.

Thanks for the input.

Could you compare the following total ideas for me? Which is better for attendance?

1) Providence v Cornell in BRIDGEPORT; St Cloud v Lowell in WORCESTER
2) Northeastern v Cornell in BRIDGEPORT; St Cloud v Lowell in WORCESTER
3) Providence v Cornell in WORCESTER; St Cloud v Lowell in BRIDGEPORT
4) Northeastern v Cornell in WORCESTER; St Cloud v Lowell in BRIDGEPORT

It's not just "Where does Providence draw best?" It's "Where does everything together draw best?"

Thanks again.

 

[FlagDUDE08]

Re: The 2014 Pairwise, Bracketology and History Thread`

It's 45 minutes to drive from Providence to Worcester and 2 hours from Providence to Bridgeport.

Not to mention both have pretty major highways going between the two roads.

 

[LynahFan]

Re: The 2014 Pairwise, Bracketology and History Thread`

Thanks for the input.

Could you compare the following total ideas for me? Which is better for attendance?

1) Providence v Cornell in BRIDGEPORT; St Cloud v Lowell in WORCESTER
2) Northeastern v Cornell in BRIDGEPORT; St Cloud v Lowell in WORCESTER
3) Providence v Cornell in WORCESTER; St Cloud v Lowell in BRIDGEPORT
4) Northeastern v Cornell in WORCESTER; St Cloud v Lowell in BRIDGEPORT

It's not just "Where does Providence draw best?" It's "Where does everything together draw best?"

Thanks again.

Maximizing overall attendance is only one consideration - the NCAA would probably rather have 4K at each regional rather than 8K at one and 2K at the other.

Having said that, my guess is that Cornell and St. Cloud draw equally in both places. I can't speak intelligently about the traveling bases of those 3 HEA schools, but since this is the internet and intelligence is not required, I'll hazard a guess that the most desirable attendance scenario would be Providence in Bridgeport and Northeastern/Lowell in Worcester.

 

[Numbers]

Re: The 2014 Pairwise, Bracketology and History Thread`

Maximizing overall attendance is only one consideration - the NCAA would probably rather have 4K at each regional rather than 8K at one and 2K at the other.

Having said that, my guess is that Cornell and St. Cloud draw equally in both places. I can't speak intelligently about the traveling bases of those 3 HEA schools, but since this is the internet and intelligence is not required, I'll hazard a guess that the most desirable attendance scenario would be Providence in Bridgeport and Northeastern/Lowell in Worcester.

Thanks Lynah. I like the way you described the hope of the NCAA for attendance. I was lacking the words to describe that clearly in my post.

And, as you can see, the current situation, if it were what the committee faced, would really not depend on St Cloud at all. Few SCSU fans would make the trip east so, the real balance is:

BC will draw great to Worcester, so we might worry about the other game, or the potential of a no-crowd final if BC were defeated in the semi. Lowell would be a good pull here, of course.

However, the Bridgeport region might have issues. Does Quinn bring fans? What about the other game there? I would assume Cornell brings a good fan base ("travels well"). Then, we are faced with this choice:

Since the games have to stay the way they are because of conference affiliation: Do Providence and Cornell draw as well to Bridgeport at Lowell would alone? And, in Worcester, Does Lowell draw there as well as Providence and Cornell would together?

And, phrased like that, I think I have answered my own questions. It seems pretty clear that StCloud/Lowell should be in Worcester and Providence/Cornell in Bridgeport.

Thanks again everyone.

 

[QUAlum2004]

Thanks Lynah. I like the way you described the hope of the NCAA for attendance. I was lacking the words to describe that clearly in my post.

And, as you can see, the current situation, if it were what the committee faced, would really not depend on St Cloud at all. Few SCSU fans would make the trip east so, the real balance is:

BC will draw great to Worcester, so we might worry about the other game, or the potential of a no-crowd final if BC were defeated in the semi. Lowell would be a good pull here, of course.

However, the Bridgeport region might have issues. Does Quinn bring fans? What about the other game there? I would assume Cornell brings a good fan base ("travels well"). Then, we are faced with this choice:

Since the games have to stay the way they are because of conference affiliation: Do Providence and Cornell draw as well to Bridgeport at Lowell would alone? And, in Worcester, Does Lowell draw there as well as Providence and Cornell would together?

And, phrased like that, I think I have answered my own questions. It seems pretty clear that StCloud/Lowell should be in Worcester and Providence/Cornell in Bridgeport.

Thanks again everyone.

Quinnipiac is 30 minutes from Bridgeport. They would send a lot of fans there along with alumni in the area. They could easily carry that region in attendence should get go all the to the regional final. They easily had to biggest support in pittsburgh at the frozen four last year.