Pairwise and Bracketology 2013 Edition

[The Exiled One]

Re: Pairwise and Bracketology 2013 Edition

If any of the number crunchers have time, could somebody calculate St. Cloud State's remaining at-large odds under the following scenarios?

MSUM win, UND win
UW win, UND win
MSUM win, CC win
UW win, CC win

Essentially, I'd like Friday morning's numbers for SCSU in advance.

 

[brassbonanza]

Re: Pairwise and Bracketology 2013 Edition

If St Cloud State misses the NCAA Tournament after winning (sharing with Minny) the toughest league in the country it might be time to "redo" the selection process. How about every league gets two bids (I know the stronger leagues hate this) with a Regular Season Champion and Playoff Champion (runner-up if same team). That should give you ten teams THEN contact MIT (just kidding) to Pairwise the remaining six teams to fill out the field. Just an idea

If the winner of the "toughest league in the country" got swept by the 5th place Hockey East and the 10th place CCHA team, perhaps it is not the toughest league in the country.

 

[The Exiled One]

Re: Pairwise and Bracketology 2013 Edition

If the winner of the "toughest league in the country" got swept by the 5th place Hockey East and the 10th place CCHA team, perhaps it is not the toughest league in the country.

This is why the other WCHA teams would like to give St. Cloud State a nice solid punch in the arm. Truth is, they'd deserve it too.

Here's the fodder: http://www.uscho.com/stats/interconference/division-i-men/2012-2013/

 

[JimDahl]

Re: Pairwise and Bracketology 2013 Edition

If any of the number crunchers have time, could somebody calculate St. Cloud State's remaining at-large odds under the following scenarios?

MSUM win, UND win
UW win, UND win
MSUM win, CC win
UW win, CC win

Essentially, I'd like Friday morning's numbers for SCSU in advance.

Start here

 

[Craig P.]

Re: Pairwise and Bracketology 2013 Edition

This would take us back to the old CC rule. CC won WCHA once upon a time and missed the tourney. That old CC rule was for the 12 team tournament and if you pulled a double (win regular and conference tourney) you would get one of the 4 byes.

You're conflating "the CC rule" with "the Clarkson rule".

The CC rule explicitly granted an automatic bid to the regular season winner of each of the four major conferences.
I'm not sure the Clarkson rule was actually on paper, but that was the one that automatically gave a bye to the winner of both the regular season and conference tournament in any of the four major conferences.

 

[brassbonanza]

Re: Pairwise and Bracketology 2013 Edition

This is why the other WCHA teams would like to give St. Cloud State a nice solid punch in the arm. Truth is, they'd deserve it too.

Here's the fodder: http://www.uscho.com/stats/interconference/division-i-men/2012-2013/

That may be, it just seems dumb to summarily state that because the team with the best conference record in the purported "toughest league" may not get an at-large, that the entire system needs to be changed, when said team has one non-UAH OOC win.

 

[Steve_MN]

Re: Pairwise and Bracketology 2013 Edition

If St Cloud State misses the NCAA Tournament after winning (sharing with Minny) the toughest league in the country it might be time to "redo" the selection process. How about every league gets two bids (I know the stronger leagues hate this) with a Regular Season Champion and Playoff Champion (runner-up if same team). That should give you ten teams THEN contact MIT (just kidding) to Pairwise the remaining six teams to fill out the field. Just an idea

And you need to recount. Given that this cannot happen this year (selection criteria already set in the handbook) the earliest it could be put in place is next year, and there will be six conferences then, allowing a total of 4 at large bids.

Hockey already has enough problems getting the tournament to get taken seriously, this would actually make it worse. At least right now, it follows the same rules as every other NCAA tournament out there, each conference gets one AQ, the rest is filled at-large.

 

[The Exiled One]

Re: Pairwise and Bracketology 2013 Edition

Start here

Perfect, thanks

To sum up for the SCSU fans...

MSUM win, UND win - 94.7% WCS
MSUM win, CC win - 89.3% WCS
UW win, UND win - 87.7% WCS
UW win, CC win - 82.3% WCS

The kicker is that the two safest scenarios are also the two that make it the most difficult to earn a #1 seed.

Go Mavs?

 

[FlagDUDE08]

Re: Pairwise and Bracketology 2013 Edition

And you need to recount. Given that this cannot happen this year (selection criteria already set in the handbook) the earliest it could be put in place is next year, and there will be six conferences then, allowing a total of 4 at large bids.

Hockey already has enough problems getting the tournament to get taken seriously, this would actually make it worse. At least right now, it follows the same rules as every other NCAA tournament out there, each conference gets one AQ, the rest is filled at-large.

Or, how about just have six bids to the tournament, one for each conference tournament winner. The two leagues with teams who last won the national championship (extrapolating out for simultaneous years) get byes, and the other four play into the Frozen Four. Simple. No hanky-panky. The only thing I might add is to require all the leagues to not allow at least one bottom team into the conference tournaments.

 

[JB]

Re: Pairwise and Bracketology 2013 Edition

You're conflating "the CC rule" with "the Clarkson rule".

The CC rule explicitly granted an automatic bid to the regular season winner of each of the four major conferences.
I'm not sure the Clarkson rule was actually on paper, but that was the one that automatically gave a bye to the winner of both the regular season and conference tournament in any of the four major conferences.

Yup, sorry about that. It was an "and" continuing a thought that inadvertently attached CC to the "double rule". An inappropriate connection as they came at different times, can't remember how many year separation. The "and" should have been a period looking back on it. Thank you for clarifying.

Funny thing is I liked the 12 team tournament better. The swapping on 5-6 seeds seemed to create a more national feel than what I (think I) see now. Always hated the bye. I still think two 3-day super regionals would be better for attendance and entertainment but I digress.

 

[Ralph Baer]

Re: Pairwise and Bracketology 2013 Edition

Or, how about just have six bids to the tournament, one for each conference tournament winner. The two leagues with teams who last won the national championship (extrapolating out for simultaneous years) get byes, and the other four play into the Frozen Four. Simple. No hanky-panky. The only thing I might add is to require all the leagues to not allow at least one bottom team into the conference tournaments.

That seems quite dumb, and how could there be "simultaneous years" -- only one team wins the championship in any year?

 

[goblue78]

Re: Pairwise and Bracketology 2013 Edition

The kicker is that the two safest scenarios are also the two that make it the most difficult to earn a #1 seed.

While you're not necessarily wrong, there is an important thing thing to understand about JimDahl's excellent work. (Correct me if I'm wrong, Jim.)These percentages are not strength weighted. Thus, while SCSU might get through in 95 percent of the scenarios, that is not the same thing as saying that SCSU has a 95 percent chance of getting through, since some of those scenarios are, by any reckoning, much more likely than others. Thus, for example, CC has a 12.5 percent rate at the moment, because they have to win three games, and in each one of them there are an equal number of scenarios in each direction, so 0.5x0.5x0.5=0.125 But CCs real chance is something closer to 3.8%, if you believe KRACH. So all the scenarios in which your team is knocked out by CC becoming an AQ are actually less likely than the scenario counts would have you believe.

 

[FreshFish]

Re: Pairwise and Bracketology 2013 Edition

And you need to recount. Given that this cannot happen this year (selection criteria already set in the handbook) the earliest it could be put in place is next year, and there will be six conferences then, allowing a total of 4 at large bids.

Hockey already has enough problems getting the tournament to get taken seriously, this would actually make it worse. At least right now, it follows the same rules as every other NCAA tournament out there, each conference gets one AQ, the rest is filled at-large.

it is up to each conference to determine how their AQ is determined. It isn't necessary to have a conference tournament. For Division I basketball, the Ivy League regular season champion gets the autobid, they don't hold a conference tournament. Until just a few years ago, the Big Ten did it the same way.

Rather than address this "issue" at the NCAA level, maybe conferences might want to re-visit how they determine who gets their autobid instead. For example, a conference could slot the regular season winner directly into the semi-finals or finals rather than force them to go through several rounds of the conference tournament. Or, you could have the regular season winner and conference tournament winner have a one-game playoff to determine the autobid (when they are different teams).

 

[JimDahl]

Re: Pairwise and Bracketology 2013 Edition

While you're not necessarily wrong, there is an important thing thing to understand about JimDahl's excellent work. (Correct me if I'm wrong, Jim.)These percentages are not strength weighted. Thus, while SCSU might get through in 95 percent of the scenarios, that is not the same thing as saying that SCSU has a 95 percent chance of getting through, since some of those scenarios are, by any reckoning, much more likely than others. Thus, for example, CC has a 12.5 percent rate at the moment, because they have to win three games, and in each one of them there are an equal number of scenarios in each direction, so 0.5x0.5x0.5=0.125 But CCs real chance is something closer to 3.8%, if you believe KRACH. So all the scenarios in which your team is knocked out by CC becoming an AQ are actually less likely than the scenario counts would have you believe.

Right. It's easy enough to strength weight them, and indeed that's what I do the rest of the year when it really makes a huge difference in the results. But a year or two ago I convinced myself that a straight share of remaining possible outcomes was more what people were looking for during the tournaments, plus it's easier to explain to people as "fact".

 

[The Exiled One]

Re: Pairwise and Bracketology 2013 Edition

While you're not necessarily wrong, there is an important thing thing to understand about JimDahl's excellent work. (Correct me if I'm wrong, Jim.)These percentages are not strength weighted. Thus, while SCSU might get through in 95 percent of the scenarios, that is not the same thing as saying that SCSU has a 95 percent chance of getting through, since some of those scenarios are, by any reckoning, much more likely than others. Thus, for example, CC has a 12.5 percent rate at the moment, because they have to win three games, and in each one of them there are an equal number of scenarios in each direction, so 0.5x0.5x0.5=0.125 But CCs real chance is something closer to 3.8%, if you believe KRACH. So all the scenarios in which your team is knocked out by CC becoming an AQ are actually less likely than the scenario counts would have you believe.

Understood

Right. It's easy enough to strength weight them, and indeed that's what I do the rest of the year when it really makes a huge difference in the results. But a year or two ago I convinced myself that a straight share of remaining possible outcomes was more what people were looking for during the tournaments, plus it's easier to explain to people as "fact".

Agreed

 

[Patman]

While you're not necessarily wrong, there is an important thing thing to understand about JimDahl's excellent work. (Correct me if I'm wrong, Jim.)These percentages are not strength weighted. Thus, while SCSU might get through in 95 percent of the scenarios, that is not the same thing as saying that SCSU has a 95 percent chance of getting through, since some of those scenarios are, by any reckoning, much more likely than others. Thus, for example, CC has a 12.5 percent rate at the moment, because they have to win three games, and in each one of them there are an equal number of scenarios in each direction, so 0.5x0.5x0.5=0.125 But CCs real chance is something closer to 3.8%, if you believe KRACH. So all the scenarios in which your team is knocked out by CC becoming an AQ are actually less likely than the scenario counts would have you believe.

If would be preferable to weight via a model... But krach can't handle ties... Which is why I prefer a score model, but its not the easiest to do w/o stat software familiarity. The only thing I can say is wait until next year for that kind of analysis. I could try to whip something up but I don't really want to.

A cheating way... Steal the results from Robin Lock's CHODR model... Which doesn't sound too bad right about now.

Personally, I prefer a modified form with offset terms for overtime... Lock eliminates all ENG and overtime goals.

 

[Umileated]

Re: Pairwise and Bracketology 2013 Edition

If would be preferable to weight via a model... But krach can't handle ties... Which is why I prefer a score model, but its not the easiest to do w/o stat software familiarity. The only thing I can say is wait until next year for that kind of analysis. I could try to whip something up but I don't really want to.

A cheating way... Steal the results from Robin Lock's CHODR model... Which doesn't sound too bad right about now.

Personally, I prefer a modified form with offset terms for overtime... Lock eliminates all ENG and overtime goals.

I have a non-iterative spreadsheet that approximates the CHODR model via fetched USCHO data. At this point in the year, the approximations aren't too far off. I'm more than happy to share it with anyone who wants it. I promised myself I'd become better with Matlab this year in order to get the iterative style working (tandem spreadsheets seemed really stupid), but here we are in the postseason...

One of the pages is a simple Poisson table where you can put in the CHODR predictions manually, and get w/l/t probabilities.

 

[Patman]

I have a non-iterative spreadsheet that approximates the CHODR model via fetched USCHO data. At this point in the year, the approximations aren't too far off. I'm more than happy to share it with anyone who wants it. I promised myself I'd become better with Matlab this year in order to get the iterative style working (tandem spreadsheets seemed really stupid), but here we are in the postseason...

One of the pages is a simple Poisson table where you can put in the CHODR predictions manually, and get w/l/t probabilities.

I can compute an entire table w a few lines of code. When I get home. Biggest challenge will be inputting it correctly.

I'd write it right now but I'm at work and using my iPhone

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Awww.. You stat and computer guys are breaking my heart. I started out wanting to be a physicist and think about elementary particles. Then life changed and I ended up in one of the caring professions. Times like this I get envious....

"I can compute and entire table in a few lines....." Envy..


"I should get out my crystal ball and just tell you what will happen"

 

[Patman]

Awww.. You stat and computer guys are breaking my heart. I started out wanting to be a physicist and think about elementary particles. Then life changed and I ended up in one of the caring professions. Times like this I get envious....

"I can compute and entire table in a few lines....." Envy..

"I should get out my crystal ball and just tell you what will happen"

I'll admit I'm bragging a bit, but these puzzles are never that easy. Otherwise it'd be a lot easier for me to do research.

With computer programming its mostly figuring out how to link things together properly.

I'll post the code for this if somebody wants a hands on. The computing "language" is freeware and supported by the academic community

---

Can I borrow your crystal ball