2013-2014 Women's D-I PairWise Contentions and Affirmations

[FiveHoleFrenzy]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

Now that this thread has rehashed yet again the traditional east vs. west brouhaha, I'm curious why people haven't picked up on the Canada vs USA gauntlet thrown down by the above-quoted post. (do we have an emoticon for ducking while brickbats begin to fly?)

Seriously, is an argument being made that eastern schools have too many Canadian players compared to western schools? I know there's a marked (and perhaps natural) difference in Canadian roster-peopling between some eastern teams near the Canadian border (e.g. Clarkson, Cornell) and some eastern teams in balmier climates (e.g. BC and Harvard).....are there differences among western schools too or is the WCHA uniformly all-American? And is an argument being made that the presence of Canadian scholar-athletes is somehow unfair, or a Bad Thing in any way? If so, in what way?

Or am I entirely missing the point of the above-quoted post?

It's pretty simple...Ignore is the best option with this poster so any so-called point is irrelevant

 

[ARM]

This is a methodologically flawed thing to do. (I assume ARM knows this but it's the kind of thing I see people do all the time so I bring it up.)

IMO, what we are looking at here is just a quick eyeball technique, not a mathematical formula. As such, it is useful to know that 9-0-1 of the WCHA's gaudy record out of conference came at Lindenwood's expense. It doesn't mean that those games have to be thrown out, but it is illustrative and something I'd want to know if I was attempting to determine just how strong the league is this year.

 

[joegrav]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

KRACH and Rutter do a better job of handling this issue.

I wasn't a math major, or even a particularly great math student, so my opinions come from the POV of a mathematics novice, but I agree with you. I like KRACH a lot, and would be pretty happy if it replaced RPI as the cornerstone of college hockey rankings.

 

[robertearle]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

Your approach makes nonconference games irrelevant as far as NCAA Tournament selection goes.

(an historical note, because I'm old. Prior to the mid 1970s, only conference champs made it into the NCAA basketball tournament; and the conference could decide if they wanted their regular season champ or their post-season champ (if they even had a post-season tournament). Some *really* good teams got left out, which made the NIT tournament a much more viable alternative to the NCAAs than it is now. Not saying that one way is better than the other - more teams is certainly better for TV revenue - just an historical note.)

 

[Eeyore]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

IMO, what we are looking at here is just a quick eyeball technique, not a mathematical formula. As such, it is useful to know that 9-0-1 of the WCHA's gaudy record out of conference came at Lindenwood's expense. It doesn't mean that those games have to be thrown out, but it is illustrative and something I'd want to know if I was attempting to determine just how strong the league is this year.

Except that it does nothing to tell you how strong the WCHA is relative to other conferences. Unless you do the same thing with all conferences and all teams of comparable strength to Lindenwood, you know less about their relative strengths than you did before you subtracted any games.

 

[ARM]

It's honestly not the worst idea, given that teams are reluctant to travel far for games due to the costs.

I'd rather go with something that is closer to the best idea than one that is just not the worst.

We could have a four-team tournament with only the conference champions as determined by each league. That would maybe be the most "fair", but on the heels of an eight-team tourney, it doesn't feel "better". That is what is done in the HS tournament, but it makes the regular season meaningless other than for seeding the conference tournaments.

Maybe we should borrow from the pokechecker index and roll up the PWR, Rutter, and KRACH into one super-ranking and use that. At least it would water down the flaws in any one ranking system.

 

[joegrav]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

(an historical note, because I'm old. Prior to the mid 1970s, only conference champs made it into the NCAA basketball tournament; and the conference could decide if they wanted their regular season champ or their post-season champ (if they even had a post-season tournament). Some *really* good teams got left out, which made the NIT tournament a much more viable alternative to the NCAAs than it is now. Not saying that one way is better than the other - more teams is certainly better for TV revenue - just an historical note.)

A worthy historical note, and much appreciated. Obviously TV revenue has driven the expansion of postseason play in all sports, college basketball included.

 

[ARM]

Except that it does nothing to tell you how strong the WCHA is relative to other conferences. Unless you do the same thing with all conferences and all teams of comparable strength to Lindenwood, you know less about their relative strengths than you did before you subtracted any games.

I didn't say they had to be subtracted and left out; it is just another data point that could easily be determined, and I believe that more data points are better, even if you still are lacking some.

 

[joegrav]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

I'd rather go with something that is closer to the best idea than one that is just not the worst.

We could have a four-team tournament with only the conference champions as determined by each league. That would maybe be the most "fair", but on the heels of an eight-team tourney, it doesn't feel "better". That is what is done in the HS tournament, but it makes the regular season meaningless other than for seeding the conference tournaments.

Maybe we should borrow from the pokechecker index and roll up the PWR, Rutter, and KRACH into one super-ranking and use that. At least it would water down the flaws in any one ranking system.

We will have to agree to disagree on such a method making the regular season meaningless. I guess it depends on how you define meaningless. As it is now, North Dakota can get whooped by Minnesota every time they play in the regular season, and in the conference tournament, but then still make the NCAA tournament, beat Minnesota by one goal on one occasion, and be crowned the champion. To me, that makes the regular season *less meaningful,* not more.

fwiw, I don't really advocate dropping to a four team tournament with conference champs only. I just think if "fairness" is the goal, that method is a lot more fair than any calculus we could ever come up with to try to objectively assign slots to at-large teams.

 

[joegrav]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

As another point of comparison, I point to the World Cup in soccer. They use some math to figure out how many teams from each continental confederation should get in, then go from there. But you could reasonably make the subjective argument that the 15th place team in Europe, which does not qualify for the World Cup, would beat the USA (#1 team in North America) in a best-of-7 series.

However: 1) the point of the World Cup is to pit teams from all over the world against one another, see how they differ, and crown a champion - not necessarily to showcase the 32 "best teams."2) tough crap, because there's a pretty clear, fair, objective standard by which, say, Sweden could have made the World Cup, and they failed to do so. So while they are probably better than the USA, they don't really "deserve" to get a crack the world championship, because we learned over a large sample that they aren't as good as some other teams in their own continent.

There's no right answer here, obviously.

 

[TonyTheTiger20]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

I'm going to play devil's advocate here for a second.

Keeping in mind that I like KRACH as much as the rest of you...

What about KRACH makes you feel that it more accurately adjusts for schedule strength than the PWR?

 

[ARM]

As it is now, North Dakota can get whooped by Minnesota every time they play in the regular season, and in the conference tournament, but then still make the NCAA tournament, beat Minnesota by one goal on one occasion, and be crowned the champion. To me, that makes the regular season *less meaningful,* not more.

Because if we say that only the conference tournament champs get into the field, and in some leagues, everybody from the conference makes the league tourney, you are just moving up that "win or go home" point. E.g., had UND won the WCHA tournament game, they'd have represented the WCHA, not Minnesota. IMO, that makes the regular season results far less meaningful.

 

[joegrav]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

Because if we say that only the conference tournament champs get into the field, and in some leagues, everybody from the conference makes the league tourney, you are just moving up that "win or go home" point. E.g., had UND won the WCHA tournament game, they'd have represented the WCHA, not Minnesota. IMO, that makes the regular season results far less meaningful.

Right. My preference would be to base it on regular season champions. Sorry for the confusion.

 

[Eeyore]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

We will have to agree to disagree on such a method making the regular season meaningless. I guess it depends on how you define meaningless. As it is now, North Dakota can get whooped by Minnesota every time they play in the regular season, and in the conference tournament, but then still make the NCAA tournament, beat Minnesota by one goal on one occasion, and be crowned the champion. To me, that makes the regular season *less meaningful,* not more.

The only way to really make the regular season more meaningful is for people to care more about it for itself. So long as the decisive arbiter of a season's success is the outcome of a postseason tournament the regular season isn't going to be terribly meaningful. There really is no method of selecting the teams to a tournament that doesn't leave something out that devalues the regular season. Even taking the regular season champs from each conference has two effects: non-conference games are meaningless and the battle for second doesn't have much value.

As I've said before, if you want the regular season to be important you need a model like European soccer leagues where the declared champion is the team that wins the regular season title.

 

[Eeyore]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

What about KRACH makes you feel that it more accurately adjusts for schedule strength than the PWR?

The most obvious thing is that KRACH doesn't rely on ad hoc corrections. Even before you get into the math details, the fact that you have to drop games from RPI to prevent wins from dragging down a teams rating is a giant, flashing red light that RPI has a serious problem. It's an admission right from the start that the method doesn't work. The committee can't use the teams' actual RPI ratings. If anyone actually believed in RPI they'd use it as is rather than going with an incomplete calculation. You can also set up very simple hypothetical situations in which RPI gives rankings that are clearly wrong.

Another advantage of KRACH that pops up before you get seriously into the math is that it is an attempt to actually model the real world. RPI just throws a bunch of numbers together without ever testing itself in any way. KRACH builds its ratings in a way that explicitly ties them to the observed winning percentages. It's not perfect but it is an actual model.

For me there's also a very strong intuitive appeal to KRACH's recognition that this is a dynamic rather than a static problem. Determining a team's quality from its results depends upon its strength of schedule. Its strength of schedule is determined by the quality of opponents. The quality of its opponents is determined by their strength of schedule. Their strength of schedule is determined by the quality of their opponents. And so on. Change one team's quality and you get ripple effects throughout the system. As I said above, RPI tries to deal with this with its third component: opponents' opponents' winning percentage. Aside from the problem that this doesn't adequately correct for the preponderance of conference games this raw measure converges to .500 too rapidly to provide meaningful information. KRACH's avoids this by having a team's rating change directly every time one of its opponents' quality changes and then using this change directly to change the rating of each of its opponents. It does this through its recursive process of running the calculations over and over again until they don't change any farther. This allows it to much more effectively capture the nature of what strength of schedule is.

There is also the fact that KRACH is based upon a statistical process, Bradley-Terry, that has proven uses in fields quite separate from college hockey. It has a proven ability to be able to rank things that do not have direct comparisons so long as there are relationships across the population being studied. RPI is a one-off toy with no theoretical justification and no applications that actually test it.

 

[TonyTheTiger20]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

The most obvious thing is that KRACH doesn't rely on ad hoc corrections. Even before you get into the math details, the fact that you have to drop games from RPI to prevent wins from dragging down a teams rating is a giant, flashing red light that RPI has a serious problem. It's an admission right from the start that the method doesn't work. The committee can't use the teams' actual RPI ratings. If anyone actually believed in RPI they'd use it as is rather than going with an incomplete calculation. You can also set up very simple hypothetical situations in which RPI gives rankings that are clearly wrong.

Another advantage of KRACH that pops up before you get seriously into the math is that it is an attempt to actually model the real world. RPI just throws a bunch of numbers together without ever testing itself in any way. KRACH builds its ratings in a way that explicitly ties them to the observed winning percentages. It's not perfect but it is an actual model.

For me there's also a very strong intuitive appeal to KRACH's recognition that this is a dynamic rather than a static problem. Determining a team's quality from its results depends upon its strength of schedule. Its strength of schedule is determined by the quality of opponents. The quality of its opponents is determined by their strength of schedule. Their strength of schedule is determined by the quality of their opponents. And so on. Change one team's quality and you get ripple effects throughout the system. As I said above, RPI tries to deal with this with its third component: opponents' opponents' winning percentage. Aside from the problem that this doesn't adequately correct for the preponderance of conference games this raw measure converges to .500 too rapidly to provide meaningful information. KRACH's avoids this by having a team's rating change directly every time one of its opponents' quality changes and then using this change directly to change the rating of each of its opponents. It does this through its recursive process of running the calculations over and over again until they don't change any farther. This allows it to much more effectively capture the nature of what strength of schedule is.

There is also the fact that KRACH is based upon a statistical process, Bradley-Terry, that has proven uses in fields quite separate from college hockey. It has a proven ability to be able to rank things that do not have direct comparisons so long as there are relationships across the population being studied. RPI is a one-off toy with no theoretical justification and no applications that actually test it.

Ah, that went down smooth. Thanks for this. That was exactly what I was looking for.

 

[Eeyore]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

The primary downside that KRACH has is that it doesn't handle the concepts of Zero and Infinity well at all. Most of the time this doesn't matter after the first few weeks of the season but last year it did. KRACH treated Minnesota not just as unbeaten but as unbeatable. North Dakota could have lost to the Gophers a hundred times and their rating wouldn't have budged.* We know that that's not true and that Minnesota had a non-zero probability of losing each game that it played. So North Dakota's KRACH rating from last year is probably slightly inflated due to this effect.

That's why having Rutter around is valuable even if you don't think it's as good as KRACH (a subject on which I really have no opinion, FWIW). It uses an entirely different statistical method that doesn't have the boundary problems that KRACH does. That's why I highlighted that UND was in the top four in both of them. Two different approaches, with different strengths and weaknesses but both clearly superior to RPI, churned out the answer that UND should have hosted a quarterfinal. Had they disagreed about that I'd be less dogmatic in my insistence that the Whioux got screwed.

As it happens, I LIKE seeing UND get screwed by the man. And when it was over it was very satisfying to see someone's career end while sitting in the box at Ridder. But I'm tired of seeing them in the first round and would prefer it if they got screwed by having to travel to play someone else this time.


*As I understand it, the USCHO calculation of KRACH adds a tie to an exactly average hypothetical team to each team's record in order to prevent an unbeaten/untied team's rating from going to infinity or a winless/tieless team's from going to zero but that this addition is not included in the strength of schedule calculation so it wouldn't affect North Dakota's rating.

 

[Nowheresville]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

As it happens, I LIKE seeing UND get screwed by the man.

I'll drink to that!

 

[CrazyDave]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

Okay, I preface this by saying that I don't have anything meaningful to add (shut up, Eeyore ).

I just had to comment that I needed to take a break just reading through the more-than-40 posts that have been added since this morning -- it took me 15 minutes to get through them all. Some were insightful, some not so much.

I'm also wondering if we're going to need a new thread -- because we'll hit the 1,000 post limit -- before the NCAA tournament field is selected. Of course, me posting this just gets us closer to that cap -- sorry.

 

[pokechecker]

Re: 2013-2014 Women's D-I PairWise Contentions and Affirmations

currently Harvard is the only non WCHA team of the top ten who has played a SOS in the top ten
and that pretty much is the case year after year
many have surmised that the WCHA 2 games against an opponent every weekend has a lot to do with the WCHA success
and that is likely part of it
but a stronger regular season schedule also makes for a stronger team
and year after year the WCHA teams dominate the top 10 SOS