Pairwise

[RSTuthill]

The HE Championship game was more or less a home game for Maine, without an actual count, I'm estimating they outnumbered us 5-1.

That said, UConn went to Orono this year and took 4 or 5 out of 6 points. IIRC.

 

[FfldCntyFan]

I’m not a hockey expert at all, but if we’re playing in Allentown, and Q and PSU are roughly equivalent, wouldn’t we prefer to play Q in front of a neutral crowd than PSU in front of a partisan one? Or is crowd support less of a factor in hockey than hoops and pigskin?

Crowd support is obviously a factor and should provide some boost for Penn St. Where ice hockey can be a bit unique is in familiarity with the rink, how the puck plays off the boards, how the ice holds up during game play and even in some cases the room in the corners below the red lines.

I'm not sure that PSU will have much more experience on that ice than the other three schools.

 

[RSTuthill]

Allentown's PPL Center is an AHL rink, so the ice, boards, glass, stanchions, doors, etc., should be quite good. Maybe facility for amplifying the bands too.

 

[DblCrossbar]

Crowd support is obviously a factor and should provide some boost for Penn St. Where ice hockey can be a bit unique is in familiarity with the rink, how the puck plays off the boards, how the ice holds up during game play and even in some cases the room in the corners below the red lines.

I'm not sure that PSU will have much more experience on that ice than the other three schools.

As for rink familiarity, that certainly didn’t help Beanpot Winner BU in the HE semis .

 

[RSTuthill]

As for rink familiarity, that certainly didn’t help Beanpot Winner BU in the HE semis .

I agree. Rinks don't make much difference at all unless they are an odd size like Wittemore last year in the Women's natty or Freitas which had half round glass mount brackets. I once saw a team dump the puck in off the half boards' glass at Freitas and it ricocheted right into the net behind the goalie who was at the top of his crease. AHA tournament game as I recall.

 

[UC15]

The HE Championship game was more or less a home game for Maine, without an actual count, I'm estimating they outnumbered us 5-1.

More like 20-1, place was Alfond South

 

[DblCrossbar]

I agree. Rinks don't make much difference at all unless they are an odd size like Wittemore last year in the Women's natty or Freitas which had half round glass mount brackets. I once saw a team dump the puck in off the half boards' glass at Freitas and it ricocheted right into the net behind the goalie who was at the top of his crease. AHA tournament game as I recall.

Whttenmore used to have an Olympic sized rink of 60 by 30 meters (197’ x 98 ft) but was changed to 200’ x 90’ (5’ wider than NHL) in 2022.

UMass at Mullen is 200’ x 95’ and is currently the widest in HE.

 

[RSTuthill]

Whttenmore used to have an Olympic sized rink of 60 by 30 meters (197’ x 98 ft) but was changed to 200’ x 90’ (5’ wider than NHL) in 2022.

UMass at Mullen is 200’ x 95’ and is currently the widest in HE.

I was at Wittemore last year for the Women's national semis as Clarkson was taking on Ohio State. The game was on a Thursday. Clarkson had beaten Minnesota, I think, to get there in a four overtime game. They were still exhausted when they got to Durham. The ice at Wittemore was not changed in size, they had just moved the boards in. Yeah, you could see that the ice was still wider than NHL. Not something you want to see when the team you are rooting for is still exhausted playing a team like Ohio State. Clarkson scored first and then it was all OSU. But OSU did not go ahead until late in the third.

 

[Waquoit]

Allentown's PPL Center is an AHL rink, so the ice, boards, glass, stanchions, doors, etc., should be quite good. Maybe facility for amplifying the bands too.

UConn has experience playing in an AHL rink.

 

[Clownshow47]

Yeah!

 

[freescooter]

Now that our season is over, I am going to once more complain about selection of the NCAA teams by a math problem. I think the pairwise is a fine metric and should be a consideration but not the only one for tournament selection. It is just a gussied up transitive property is sports which misses lots of nuances. There are 16 slots. 5 or is it 6 are awarded to league tournament champs. When there were only a few slots it made sense maybe but with 10 it is time to stop pretending that the best teams get there through solving a math problem. The NCAA Frozen Four is a 1,2,3,and4 seed. That suggests that the magic formula did a pretty random job of ranking team.

 

[Patman]

Now that our season is over, I am going to once more complain about selection of the NCAA teams by a math problem. I think the pairwise is a fine metric and should be a consideration but not the only one for tournament selection. It is just a gussied up transitive property is sports which misses lots of nuances. There are 16 slots. 5 or is it 6 are awarded to league tournament champs. When there were only a few slots it made sense maybe but with 10 it is time to stop pretending that the best teams get there through solving a math problem. The NCAA Frozen Four is a 1,2,3,and4 seed. That suggests that the magic formula did a pretty random job of ranking team.

The real problem is it's basically the RPI formulation with extra steps.

Also you have to realize this is hockey. Go to CHN and look up their "power ratings". The model form is just logistic regression without home road factory. You'll see that while RPI doesn't do a great job the results won't be too far off in general.

---

There's a reason why baseball does double elimination. If hockey wasn't so physically intensive I'd suggest the same.

There really aren't great options for a mostly fair format.

 

[SubbaBub]

Hockey isn't FB or even BB. If you put the 16 best teams into a hat, any of them can win 2 games with the possible exception of the Atlantic champion in most years.

Cornell was a 4 seed and well outside the top 16 but they play in the ECAC, home of many former championship programs, they can certainly beat top squads and did.

The pairwise is vital to a 16 team field so undeserving power programs are penciled in every year based on reputation or number of pro prospects. For hockey, it works and works well.

As for the seeding and site assignment, seeds don't matter much beyond the top 8, they should get an easier first game even though it isn't so easy. Attendance still matters for this sport. Regional sites are important as is sending local teams to them. If your team can't handle and thrive a hostile barn, are you even a hockey player?

 

[RSTuthill]

Now that our season is over, I am going to once more complain about selection of the NCAA teams by a math problem. I think the pairwise is a fine metric and should be a consideration but not the only one for tournament selection. It is just a gussied up transitive property is sports which misses lots of nuances. There are 16 slots. 5 or is it 6 are awarded to league tournament champs. When there were only a few slots it made sense maybe but with 10 it is time to stop pretending that the best teams get there through solving a math problem. The NCAA Frozen Four is a 1,2,3,and4 seed. That suggests that the magic formula did a pretty random job of ranking team.

The thing about hockey is that the results are so heavily influenced by goalies that are on hot streaks. PSU, for instance was not great in the first half of the season. In the second half they were on fire and Sergeev undoubted had something to do with that. That in 28 shots UConn could not get the winning goal on Sunday speaks to that. And remember that the entire season is used in team selection.

In WBB, there are three #1's and one #2 left. If you look at the Vegas odds, the #2 is heavily favored to take it all.

I have followed college hockey ever since I was an undergrad at Clarkson in the 60's. They just got into a smoke filled room and the selection committee argued. In those days the Ivy League had a lot of stroke and their champion often bumped out a more deserving team in the east. I never want to go back to those days. Nor do I want any subjectivity at all allowed back into the process.

 

[freescooter]

But here’s the difference between then and now. In the 60s and well into 70s There were only 4 teams. Into the mid 80s it fluctuated between 5-8. And there were only a couple of leagues. It was a totally different world. Each league basically got together and picked its “best” team, ie the one it thought had the best chance to win. So if UNH had the best record but had been on a losing streak but Maine was 3rd but had won 5 straight the Black Bears might get the nod. At one point in fact each league got 1 bid for its tourney.champ and 1 to select however it wanted. The NCAA didn’t decide that at all. Now there are 10 at-large bids. So the Ivy champ isn’t bumping Clarkson. From 1948-early 1970s the ECAC picked 2 teams and the WCHA pick its 2. The NCAA was barely involved other than sponsoring the tournament. When the ccha came on the scene, the tournament went to 5 then 6 but the leagues still chose the teams, not the NCAA. When Hockey East came along it eventually went to 8. Each league got 2 automatic bids. One to the league tournament champ and one to whoever it wanted. Usually if they lost the tourney, the regular season champ, but not always. My recollection is that when the tournament went to 12 that system stayed in place but a couple of at-larges were added. When the late lamented CHA and the MAAC/AHA came along selection changed again, it was these 2 interlopers, not the established leagues that got screwed. The old guard leagues invented rules to delay giving bids to the newcomers. They extended the time line a league needed to play before it got an automatic bid. Then they invented out of whole cloth the “insularity” rule claiming the MAAC teams didn’t play enough non-conference games despite the majors being unwilling to schedule them. In any case no such rule was on the books. They just made it up. One year for some quirky schedule reason BC or BU , I think, actually played more games against other HEA opponents than any MAAC team did in its league. Theninsulsr schedule rule was determined not to apply for some reason. LOL.

I agree with you that a hot goalie can have a huge impact. Like a pitcher in baseball. But I don’t agree that current performance ought not be weighed at least somewhat. In theory, Penn State’s season could have gone in reverse. Piling up wins early then struggled late but still earning enough “points” in PWR to get into the tournament. I just think other factors deserve some weight. Most of the time it will only matter at the margins. I just think the current system makes assumption that aren’t any more accurate than a system that weights other factors would. For all the complex math involved it is nothing more than, as I said, a gussied up version of the transitive property applied the sports. If UConn beats BU and BU beats Denver than UConn is better than Denver. This years Frozen Four is a perfect example of how the glorified transitive property doesn’t predict anything. The NCAA tournament on the other hand both mens and womens, with 4 #1s and 3 ones and a 2 is much more what would be expected. And by the way, UConn was a 2 seed because of its NET which is basketballs PWR. Based on most metrics it was worthy of a 1 seed.

 

[RSTuthill]

But here’s the difference between then and now. In the 60s and well into 70s There were only 4 teams. Into the mid 80s it fluctuated between 5-8. And there were only a couple of leagues. It was a totally different world. Each league basically got together and picked its “best” team, ie the one it thought had the best chance to win. So if UNH had the best record but had been on a losing streak but Maine was 3rd but had won 5 straight the Black Bears might get the nod. At one point in fact each league got 1 bid for its tourney.champ and 1 to select however it wanted. The NCAA didn’t decide that at all. Now there are 10 at-large bids. So the Ivy champ isn’t bumping Clarkson. From 1948-early 1970s the ECAC picked 2 teams and the WCHA pick its 2. The NCAA was barely involved other than sponsoring the tournament. When the ccha came on the scene, the tournament went to 5 then 6 but the leagues still chose the teams, not the NCAA. When Hockey East came along it eventually went to 8. Each league got 2 automatic bids. One to the league tournament champ and one to whoever it wanted. Usually if they lost the tourney, the regular season champ, but not always. My recollection is that when the tournament went to 12 that system stayed in place but a couple of at-larges were added. When the late lamented CHA and the MAAC/AHA came along selection changed again, it was these 2 interlopers, not the established leagues that got screwed. The old guard leagues invented rules to delay giving bids to the newcomers. They extended the time line a league needed to play before it got an automatic bid. Then they invented out of whole cloth the “insularity” rule claiming the MAAC teams didn’t play enough non-conference games despite the majors being unwilling to schedule them. In any case no such rule was on the books. They just made it up. One year for some quirky schedule reason BC or BU , I think, actually played more games against other HEA opponents than any MAAC team did in its league. Theninsulsr schedule rule was determined not to apply for some reason. LOL.

I agree with you that a hot goalie can have a huge impact. Like a pitcher in baseball. But I don’t agree that current performance ought not be weighed at least somewhat. In theory, Penn State’s season could have gone in reverse. Piling up wins early then struggled late but still earning enough “points” in PWR to get into the tournament. I just think other factors deserve some weight. Most of the time it will only matter at the margins. I just think the current system makes assumption that aren’t any more accurate than a system that weights other factors would. For all the complex math involved it is nothing more than, as I said, a gussied up version of the transitive property applied the sports. If UConn beats BU and BU beats Denver than UConn is better than Denver. This years Frozen Four is a perfect example of how the glorified transitive property doesn’t predict anything. The NCAA tournament on the other hand both mens and womens, with 4 #1s and 3 ones and a 2 is much more what would be expected. And by the way, UConn was a 2 seed because of its NET which is basketballs PWR. Based on most metrics it was worthy of a 1 seed.

I am amenable to a weighting factor that gradually increases linearly as the season progresses. That should be easy to implement in non dimensional form. Say start at 1 and increases to two at the end for the last game of the RS and stays at 2 during the conference tournaments. In other words the result of the last game of the RS is twice as important as the result of the first game. Something like that.

 

[zls44]

OR - we could play more non-conference later in the season. Those are the games that really make the PWR move a lot, and all of them are in the first half.

 

[RSTuthill]

OR - we could play more non-conference later in the season. Those are the games that really make the PWR move a lot, and all of them are in the first half.

Interesting idea, but the early games are when teams are trying to optimize their teams and I doubt they would want to start conference play right away.

 

[Patman]

"Last x games" factor in the RPI in the past was removed because it tended to punish the better conferences because if you're only playing each other somebody has to lose.

I wouldn't mind it as such in something like KRACH (weighted bradley terry) while it stops it being a predictive model per se (and frankly it can't be with ties, anyhow) it at least can balance on opponent strength

 

[Patman]

OR - we could play more non-conference later in the season. Those are the games that really make the PWR move a lot, and all of them are in the first half.

While OOC will move the needle more i don't think that either helps or hurts a team

If we went with the old "last 16" I could see a final weekend against a pair of Atlantic hockey opponents to try to goose the ratings... but hockey coaches tend to be more cultural than pragmatic and such nonconf games are poor ticket sellers

 

[SubbaBub]

All games should count equally. Hockey East has its own late season PWR booster. It's called league play and the HET. UConn moved up 6 spots in the rankings over the last month of the season.

There really isn't a problem here, and if there is, it benefits us.

 

[huskymedic]

-> Pairwise Rankings on way out​

The Pairwise Rankings, which have long been used to determine the NCAA tournament field, are likely on their way out.

The Pairwise Rankings are expected to be replaced by the NCAA Percentage Index (NPI), which is used on the women's side.

Like the Pairwise, the NPI will have weights for different criteria such as winning percentage, strength of schedule, home-road bonuses, quality win bonuses and overtime wins.

The coaches were presented a document that showed last year's final Pairwise Rankings alongside how the NPI would have looked at the end of the season, depending on different weights.

The NCAA tournament field would have looked relatively similar. In a couple of NPI formulas, Arizona State would have replaced Penn State as the final team in the tournament.

The NPI will likely be used beginning this season. <-

 

[RSTuthill]

Comment, Patman? Not KRACH, but a little more sophisticated than the Pairwise.

 

[freescooter]

Comment, Patman? Not KRACH, but a little more sophisticated than the Pairwise.

You know what the most sophisticated way to rank teams is? W-L-T. Everything else is an attempt to find distinctions where there are no differences and differences where there are no distinctions. It’s why exactly zero professional leagues use these ginned up formulas for post season access.

 

[RSTuthill]

You know what the most sophisticated way to rank teams is? W-L-T. Everything else is an attempt to find distinctions where there are no differences and differences where there are no distinctions. It’s why exactly zero professional leagues use these ginned up formulas for post season access.

There is a big difference. Professional leagues play a lot more games with a lot of interlocking scheduling. In college sports there are many more teams and a lot fewer games with much less interlocking of schedules. Also some conferences are historically strong, others not so much. And pros can afford a lot more travel because of all the revenue and time spent exclusively on that sport.