Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

I'm not saying there is a right or wrong answer but perhaps some detail would contribute more to the thread here

I also doubt it. Though Mercyhurst deserved better than Wisconsin when going strictly by the PWR, the 2012 criteria definitely delivered mixed messages. When the formal criteria were conflicting, I think the committee clearly exercised its discretion. based on conference tournament results and travel costs. It'd be wrong to conclude that the committee always does whatever it wants to save money. There are tradeoffs. I don't think the committee would go against all the criteria to save money on Mercyhurst vs BU/BC as opposed to Cornell.

At the moment, Harvard and Clarkson are ahead of Mercyhurst in all criteria, except the H2H split vs. Clarkson. So if season ended today with the current standings, faced with the tradeoff of saving money sending Mercyhurst to Cornell instead BU/BC, vs. going against the criteria in assessing Mercyhurst relative to Harvard/Clarkson, I don't anticipate they'd ever send Mercyhurst to Cornell. I don't know any precedent that would lead me to expect that outcome.

 

[TonyTheTiger20]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Thatttt's what I was looking for.

Interesting. Why are the criteria so ambiguous again?

 

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Thatttt's what I was looking for.

Interesting. Why are the criteria so ambiguous again?

Well, the coaches asked the same question last year. Some would definitely prefer the system to be very rigid like the men's. I don't know what the committee decided in the offseason, and unlike past years, the meeting minutes aren't readily available on the NCAA site. (EDIT: nor are the championship handbooks which are another source -- they seem to have stopped posting them online for all sports, the latest available being from 2011. It's also possible if they went to a more rigid criteria, they wouldn't actually legislate it and it wouldn't show up in the meetings or handbooks anyway, since a more rigid criteria would not conflict with the current rules on the books).

Of course one huge obstacle to having a more rigid system is that the men's bracket is fully seeded, while the women are half-seeded and at the mercy of travel considerations. Yet last year's pairing of SLU with BC and Mercyhurst with Wisconsin wasn't just about travel.

It's possible the committeee would stick closer to the PWR, barring travel cost issues. Personally, I think that would be a mistake. I like the policy that you're rigid when there's strong consensus among the criteria, but then you leave it more up to the committee's discretion the more there's considerable conflict among the criteria. I find that to be a happy medium between the overly rigid men's hockey system and the much more subjective basketball system which is horribly biased towards the BCS schools.

 

[OnMAA]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

So if let's say NE wins HE, a more likely scenario than a team other than Harvard, Clarkson or Cornell winning the ECAC, you could end up with only one WCHA team in the tournament. Don't think anyone would have predicted that...

No one might have predicted it, but it would be wildly celebrated by 85% of the geography that takes interest!

With NE (or is it NU ?) winning Beanpot, the musings from a little while ago all of a sudden look more realistic...or as IceIsNice would say...More Promising.

(Although He selfishly admitted that he is more interested in MC making a trip to Bean Town in a few weeks. )

 

[Watson Rink]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Holy cow, the Beanpot is the Great Leveller: #10 NU prevails over #2 BC and #3 BU while #5 Harvard pulls even for a season split with #3 BU. Next step is for NU to win the HEA autobid the way they just won the Beanpot. Next step is for the three leading ECAC teams to come out with a two-way tie in the regular season, then have the ECAC #3 seed surprise by winning the autobid. Penultimate step is for the non-Minnesota teams to do likewise by achieving parity in the remaining WHCA regular season and tournament. Final step is for Mercyhurst to finish like gangbusters. Then you would have Minnesota plus two quite plausible but non-top eight autobid winners plus two top-eight HEA teams, two top-eight ECAC teams, two top-eight WCHA teams and top-eight Mercyhurst, resulting in two top-eight teams being squeezed out of the NCAAs by the autobids. The amazing thing is....it just might happen that way!

 

[IceIsNice]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

With NE (or is it NU ?) winning Beanpot, the musings from a little while ago all of a sudden look more realistic...or as IceIsNice would say...More Promising.

(Although He selfishly admitted that he is more interested in MC making a trip to Bean Town in a few weeks. )

Honestly at this point, I just hope that MC.... (now MU, actually), is still playing hockey in mid-March. Ugh!

 

[CrazyDave]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Putting this here too.

NCAA Frozen Four full-session ticket prices increase $5 on Sunday (Feb. 17) so get your tickets today or tomorrow by going to https://www.mygophersports.com. There are also some promotional packages available.

Single-session tickets won't be available until the week of the tournament (starting March 18).

 

[Stat Man]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Holy cow, the Beanpot is the Great Leveller: #10 NU prevails over #2 BC and #3 BU while #5 Harvard pulls even for a season split with #3 BU. Next step is for NU to win the HEA autobid the way they just won the Beanpot. Next step is for the three leading ECAC teams to come out with a two-way tie in the regular season, then have the ECAC #3 seed surprise by winning the autobid. Penultimate step is for the non-Minnesota teams to do likewise by achieving parity in the remaining WHCA regular season and tournament. Final step is for Mercyhurst to finish like gangbusters. Then you would have Minnesota plus two quite plausible but non-top eight autobid winners plus two top-eight HEA teams, two top-eight ECAC teams, two top-eight WCHA teams and top-eight Mercyhurst, resulting in two top-eight teams being squeezed out of the NCAAs by the autobids. The amazing thing is....it just might happen that way!

Problem for Mercyhurst is how do you finish like gangbusters when you only have Syracuse, RIT and CHA playoff left on your schedule. When 12 of your games are against RIT, Penn and Lindenwood then you would need a really strong out of conference schedule to make up for that. I don't think they did themselves any favors by playing Yale, MSU and Providence as 6 of the out of conference games.

Even if they run the table they may well find themselves on the outside looking in come tournament time.

 

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Problem for Mercyhurst is how do you finish like gangbusters when you only have Syracuse, RIT and CHA playoff left on your schedule. When 12 of your games are against RIT, Penn and Lindenwood then you would need a really strong out of conference schedule to make up for that. I don't think they did themselves any favors by playing Yale, MSU and Providence as 6 of the out of conference games.

Even if they run the table they may well find themselves on the outside looking in come tournament time.

I don't buy the criticism. Mercyhurst is a solid #7 right now. Yes, Mercyhurst would be screwed if there were two non-top 8 autobid winners, but so what? The claim is that if Mercyhurst had played a different schedule, the team would have a good chance of moving ahead of Clarkson and Cornell? They played those two teams. They went 1-2 against them and 1-1 against BC. If they wanted to be ahead of Clarkson and Cornell, they should've gone 2-1 against them. I think they made out well being #7 based on the results they have to date. Besides it's never happened that a solid #7 in the PWR got knocked out by two non-top 8autobids (but I agree it's as likely to happen this year as any.)

As for the specifics on the schedule: playing Minnesota State was actually great for Mercyhurst, because they swept them and the WCHA rivals for at-large berths didn't. I actually think it's good they finish in the CHA schedule, at least for NCAA purposes, because I think the downside risk in losing TUC games (were they to play those games) and falling out of the #7 spot, is greater than the upside of possibly moving up to #6 were they to play a tougher schedule.

 

[Eeyore]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

As for the specifics on the schedule: playing Minnesota State was actually great for Mercyhurst, because they swept them and the WCHA rivals for at-large berths didn't. I actually think it's good they finish in the CHA schedule, at least for NCAA purposes, because I think the downside risk in losing TUC games (were they to play those games) and falling out of the #7 spot, is greater than the upside of possibly moving up to #6 were they to play a tougher schedule.

Agreed, and I think this highlights some of the serious flaws with PWR as a ranking system.

 

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Agreed, and I think this highlights some of the serious flaws with PWR as a ranking system.

Yes and no. I agree it's flawed too, but I wouldn't have a horrible problem with Mercyhurst getting in over North Dakota or Wisconsin right now.

I do agree the record vs. TUC criteria is deeply flawed as a measure of "success against better teams." In a wider range of measures of that ideal, Mercyhurst would be on par with North Dakota and Wisconsin, not way way ahead.

But if the final outcome is one where the stats that consider the whole season are very close, and the stats that consider results against the best teams on very close, is it a horrible injustice if you then take the team that had the better record vs. common opponents (even if they played these teams a different number of times)? I think there's an appeal in using something transparent and easy to interpret like H2H and common opponents when other measures are all really close.

Right now UND and Wisconsin are slightly ahead of Mercyhurst in KRACH and the polls, but Mercyhurst is ahead in RPI. If we reached the point where KRACH gave a huge edge to UND and Wisconsin, then I'd think, as usual, the WCHA was being screwed by using a flawed an arbitrary method like the RPI. But with the current set of results, I don't see a huge problem with the PWR taking Mercyhurst.

 

[Eeyore]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Yes and no. I agree it's flawed too, but I wouldn't have a horrible problem with Mercyhurst getting in over North Dakota or Wisconsin right now.

I don't really intend to be making a comment about which of the teams is more deserving, at least not right now. It's more that I have a bachelor's degree in statistics and PWR offends my mathematical sensibilities on a level beyond just wondering which team it treats unfairly.

 

[ARM]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

RPI is just horribly overmatched for this question; note the love it gives to Holy Cross and St. Anselm.

What is odd to me this year is how much KRACH and Rutter differ. In Rutter, North Dakota is far ahead of Mercyhurst, considerably better than Clarkson, and right on the heels of Harvard. That's likely because Rutter is the least forgiving of losses to weaker teams. UND did lose to MSU, but it has three wins to balance that out, whereas the Lakers are only 1 of 4 against RMU and Clarkson is just .500 with both Colgate and Princeton. Rutter is also likely better able to discount UND's four losses to Minnesota, while a measure like RPI can only compute those as losses and has no knowledge as to where those losses were incurred.

 

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

What is odd to me this year is how much KRACH and Rutter differ. In Rutter, North Dakota is far ahead of Mercyhurst, considerably better than Clarkson, and right on the heels of Harvard. That's likely because Rutter is the least forgiving of losses to weaker teams. UND did lose to MSU, but it has three wins to balance that out, whereas the Lakers are only 1 of 4 against RMU and Clarkson is just .500 with both Colgate and Princeton. Rutter is also likely better able to discount UND's four losses to Minnesota, while a measure like RPI can only compute those as losses and has no knowledge as to where those losses were incurred.

Thanks for pointing that out. I hadn't been following Rutter lately. Your guess is about right, but I think the difference isn't about "Rutter (generally) being the least forgiving of losses to weaker teams" which is more a statement about the distributional assumptions of the two rankings -- Rutter uses normal distributions while KRACH uses logit distributions, and logit has fatter tails, so KRACH is usually not as sensitive to extreme events like losing to a bad team.

I think the difference here is much more about how the two methods handle undefeated Minnesota. Rutter's able to determine Minnesota's coefficient by using Bayesian methods, i.e. assuming a prior distribution and allowing results to inform the prior. The USCHO KRACH I believe just gets around this by giving everyone a fake tie to an average team. Rutter says ND's chance of win vs. Minnesota is 1 of 200 (0.5%), while KRACH says it's 3% (about twice Minnesota's fake loss percentage of 0.5/31.5).

To see how this difference influences the rankings, if it turned out that Rutter was accurate about the probability of ND beating Minnesota, and Minnesota's KRACH rating were accurate, then ND's KRACH should be about 600, on par with Harvard, and ND is about on par with Harvard in Rutter. Also, the North Dakota probability of beating Clarkson is about 60% in Rutter, and it would be about 60% in KRACH if North Dakota's KRACH was 600. So it seems like the difference in punishment ND & Wisconsin get for losing to Minnesota might actually matter a lot between Rutter and KRACH.

 

[northhockey]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

UND drops to 10th in PairWise after their sweep of OSU. Really

 

[OnMAA]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

UND drops to 10th in PairWise after their sweep of OSU. Really

Don't know what the list of the top 12 was beforfe the weekend, but UND dropped probably because OSU or UMD are no longer in the top 12, so those wins vs OSU now mean much less in the pairwise rankings, as they now only count towards the RPI number, but no longer in the HOH or TUC comparisons. Suspect that UND compares less favorably vs. another team that crept into the top 12, or another team improved their TUC and HOH as a result of a different team entering the top 12. Me thinks it is the SLU/Darty entries in spots 11 and 12 helping NE move up the ladder. So the irony is that UND wins over OSU may hurt UND in the pairwise by knocking OSU out of the top 12. They can hurt themselves even more by keeping UMD out of the top 12, by winning those games next weekend.

 

[dave1381]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

UND drops to 10th in PairWise after their sweep of OSU. Really

Teams 11 to 14 are so close in the RPI. There will be lots of volatility based on who is in and out.

I don't think there's anything wrong with Wisconsin passing UND by a hair after sweeping UMD. UND has to at least beat Wisconsin in RPI to beat out a team it lost to 3 of 4 times.

The Northeastern vs. North Dakota comparison is a tricky one. At first I thought it was silly that Northeastern was ahead, but Northeastern does have the better results against the better teams, and the common opponent comparison is based on Clarkson, not some random lower tier team.

Now the sad thing is North Dakota now plays UMD, so UMD is probably out of the RPI top 12 unless UND loses to them! Truly this is the worst feature of the criteria.

But in the end, I don't expect the UND Northeastern comparison will matter at all, because Wisconsin has the comparison with Northeastern pretty much locked up. UND has to pass Wisconsin, and it'll likely be about who has the better WCHA tournament. I don't expect a controversial outcome if the last bid comes down to these three teams.

 

[brookyone]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

So the irony is that UND wins over OSU may hurt UND in the pairwise by knocking OSU out of the top 12. They can hurt themselves even more by keeping UMD out of the top 12, by winning those games next weekend.

Rather than simple irony, I'm going to go ahead and consider it a full fledged and blatant flaw in the formula.

 

[Eeyore]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

The Northeastern vs. North Dakota comparison is a tricky one. At first I thought it was silly that Northeastern was ahead, but Northeastern does have the better results against the better teams, and the common opponent comparison is based on Clarkson, not some random lower tier team.

The problem isn't who the games were against. It's that the comparison based on a total of three games: Northeastern played Clarkson once and UND played them twice. With that small a sample size, it really has no meaning.

Now the sad thing is North Dakota now plays UMD, so UMD is probably out of the RPI top 12 unless UND loses to them! Truly this is the worst feature of the criteria.

Bad, but it's not even close to the worst feature of the common opponents criterion. Until they normalize for the number of games played against each opponent, that's the biggest flaw.

 

[TonyTheTiger20]

Re: Preliminary & Ongoing Pairwise Watch and 2013 NCAA Tournament Musing Thread

Didn't they just make a change to CoOpp record on the men's side?

I think the way it works now is:

Team A:
30-0 vs. Lindenwood
0-1 vs. Minnesota

Team B:
1-0 vs. Lindenwood
0-2 vs. Minnesota

On the women's side team A wins this comparison 30-1 to 1-2.

With the men's rules I believe they say "Both teams have a 1.000 win vs Lindenwood and .000 win % vs. Minnesota, so they tie the comparison.

Someone correct me if I'm wrong.