Pairwise and Bracketology 2013 Edition

[UnionHockeyManiac]

Re: Pairwise and Bracketology 2013 Edition

Which teams are 100% in right now? For the teams that are not 100%, can someone put approximate percentages on all other teams who have a chance?

 

[LTsatch]

Re: Pairwise and Bracketology 2013 Edition

Which teams are 100% in right now? For the teams that are not 100%, can someone put approximate percentages on all other teams who have a chance?

From reading what the experts have to say on here, you will most likely have to wait until after this weekend to get those percentages, too many variables at this point.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

From reading what the experts have to say on here, you will most likely have to wait until after this weekend to get those percentages, too many variables at this point.

LT is correct. And, this is why:
Remember the 4 pieces of each comparison. RPI, TUC, ComOpp, H2H. At this point, there is likely to be only slight movement in the CommOpp and H2H part of any particular comparison.

However, a team's RPI can move perhaps +/- .100 this weekend, and, in the case of ECAC teams, the same next weekend if they get that far. That has 2 effects: It affects how their RPI stacks up against other teams, and it affects which teams are TUCs.

There are a few schools whose TUC status apparently makes a big difference in the at-large field. Among these are
Michigan (because of Notre Dame)
Merrimack (because of BU)
Colorado College (because of Wisconsin)

That being said, Priceless has remarked that it seems that
Quinnipiac and Minnesota are at this point locks for the field.

I assume the poster from 2 posts up is a Union backer. Without doing a full up recreation of this weekend, I would say that Union has to win this weekend, or it's hard for them, because Dartmouth would likely pass them in the PWR. And, it could be that they also will have to not lose 2 games next weekend. But, that's sort of a "Having looked at this many times before, this is my guess" calculation.

 

[Patman]

Re: Pairwise and Bracketology 2013 Edition

The reality is from Lowell on down it is bunched up in that a couple of losses can lead to a significant fall. Some teams are more proofed than others but it may depend on the minutiae of other results.

 

[kingdobbs]

Re: Pairwise and Bracketology 2013 Edition

LT is correct. And, this is why:
Remember the 4 pieces of each comparison. RPI, TUC, ComOpp, H2H. At this point, there is likely to be only slight movement in the CommOpp and H2H part of any particular comparison.

However, a team's RPI can move perhaps +/- .100 this weekend, and, in the case of ECAC teams, the same next weekend if they get that far. That has 2 effects: It affects how their RPI stacks up against other teams, and it affects which teams are TUCs.

There are a few schools whose TUC status apparently makes a big difference in the at-large field. Among these are
Michigan (because of Notre Dame)
Merrimack (because of BU)
Colorado College (because of Wisconsin)

UConn as well. UConn becoming a TUC gives Niagara *just enough* games vs. TUCs to turn some comparisons from null, and a 60% win percentage is enough to win a sufficient number to insulate the Purps against losing the league tourney.

 

[Priceless]

Re: Pairwise and Bracketology 2013 Edition

Quinnipiac is the overall #1 seed, will be #1 in Providence.
Minnesota has clinched, can only be knocked from a 1-seed by a weird confluence of events, and will likely be the 1-seed in Grand Rapids.
Miami has clinched and will likely be the 1-seed in Toledo.

Everyone else still has work to do. UND and Denver can clinch a spot by winning this weekend. Lowell, Boston College and New Hampshire would clinch with wins. New Hampshire is very close but plays TUC Providence this weekend on the road which really hurts them if they are swept. I don't think anyone else from the CCHA or ECAC can clinch this weekend.

I went back to look at last year and only Boston College had clinched a spot before the conference quarters. A bunch of teams were able to clinch by winning those series. In fact, the field was pretty much set last year, barring what we referred to as "the parley" with a bunch of upset winners in the conferences (Bowling Green, Providence, Harvard etc). What was still to be determined was the final order of the teams.

I also found this nugget, which came to mind when the DIY calculator put UNH 4th and UND 5th in an earlier computation:

I was looking back through the archives last night and came upon this contradiction. The top seeds in 2005 were Boston College, Minnesota (host), Denver and CC. There was a 3-way tie for 11th between BU, Maine and Wisconsin. Were this three way tie broken by RPI then the result would have been:
11 Boston Univ (HE) 16 0.5685
12 Maine (HE) 16 0.5607
13 Wisconsin (WC) 16 0.5537

Maine and BU would be #3 seeds and Wisconsin would have been a #4. Boston College would have been forced to play Wisconsin to avoid a WCHA game in round one. Instead, that tie was broken:
11 Wisconsin (WC) 16 0.5537
12 Boston Univ (HE) 16 0.5685
13 Maine (HE) 16 0.5607

The Badgers won the comparisons vs Maine and Boston University, despite having the lowest RPI of the three. BU won the comparison with Maine and was seeded 12th. Wisconsin became a #3 seed and an all-WCHA game was averted as Maine dropped to #4 and went to Minnesota.

Contrast that with 2007 when St Lawrence, UMass and Maine were again tied at 11th place. Had the committee followed precedent the tie would have been broken:
11 Mass-Amherst (HE) 13 0.5328
12 St Lawrence (EC) 13 0.5384
13 Maine (HE) 13 0.5366

as the Minutemen won comparisons with both SLU and Maine. SLU then won the comparison with Maine. However, they chose to break the tie by RPI:
11 St Lawrence (EC) 13 0.5384
12 Maine (HE) 13 0.5366
13 Mass-Amherst (HE) 13 0.5328

So UMA was the #4 seed that played Clarkson while Maine was a #3 seed and played St. Cloud.

We have been operating under the belief that the NCAA would break ties involving three teams or more by RPI, however the 2005 case shows that they also break ties using comparisons. The same logic can hold true with ties of two teams. The NCAA has not clarified how ties are broken. More than likely, it is so they can avoid a case like 2005 when they would have either sent #4 Wisconsin to #1 Minnesota or #1 Boston College. This gives them added flexibility at the expense of transparency.

 

[unh1982]

Re: Pairwise and Bracketology 2013 Edition

Quinnipiac is the overall #1 seed, will be #1 in Providence.
Minnesota has clinched, can only be knocked from a 1-seed by a weird confluence of events, and will likely be the 1-seed in Grand Rapids.
Miami has clinched and will likely be the 1-seed in Toledo.

Everyone else still has work to do. UND and Denver can clinch a spot by winning this weekend. Lowell, Boston College and New Hampshire would clinch with wins. New Hampshire is very close but plays TUC Providence this weekend on the road which really hurts them if they are swept. I don't think anyone else from the CCHA or ECAC can clinch this weekend.

I went back to look at last year and only Boston College had clinched a spot before the conference quarters. A bunch of teams were able to clinch by winning those series. In fact, the field was pretty much set last year, barring what we referred to as "the parley" with a bunch of upset winners in the conferences (Bowling Green, Providence, Harvard etc). What was still to be determined was the final order of the teams.

I also found this nugget, which came to mind when the DIY calculator put UNH 4th and UND 5th in an earlier computation:

Unh can also move up just as fast , a sweep moves them up to 4th in the power.... putting them back in control of there destiny , which we all know means %^#$.....lol

 

[Patman]

Unh can also move up just as fast , a sweep moves them up to 4th in the power.... putting them back in control of there destiny , which we all know means %^#$.....lol

Assuming Lowell doesn't do the same... Here's the thing... Middle of the pack usually behaves reliably since there are so many of them... However, at the top things get thinner and are more co-dependent on a fewer number of results.

 

[MarkEagleUSA]

Re: Pairwise and Bracketology 2013 Edition

Quinnipiac is the overall #1 seed, will be #1 in Providence.

Is this a a given regardless of what happens over the next 2 weekends?

 

[unh1982]

Re: Pairwise and Bracketology 2013 Edition

Assuming Lowell doesn't do the same... Here's the thing... Middle of the pack usually behaves reliably since there are so many of them... However, at the top things get thinner and are more co-dependent on a fewer number of results.

lowell doesnt move up , not playing a tuc , Had them sweeping maine .

 

[UnionHockeyManiac]

Re: Pairwise and Bracketology 2013 Edition

LT is correct. And, this is why:
Remember the 4 pieces of each comparison. RPI, TUC, ComOpp, H2H. At this point, there is likely to be only slight movement in the CommOpp and H2H part of any particular comparison.

However, a team's RPI can move perhaps +/- .100 this weekend, and, in the case of ECAC teams, the same next weekend if they get that far. That has 2 effects: It affects how their RPI stacks up against other teams, and it affects which teams are TUCs.

There are a few schools whose TUC status apparently makes a big difference in the at-large field. Among these are
Michigan (because of Notre Dame)
Merrimack (because of BU)
Colorado College (because of Wisconsin)

That being said, Priceless has remarked that it seems that
Quinnipiac and Minnesota are at this point locks for the field.

I assume the poster from 2 posts up is a Union backer. Without doing a full up recreation of this weekend, I would say that Union has to win this weekend, or it's hard for them, because Dartmouth would likely pass them in the PWR. And, it could be that they also will have to not lose 2 games next weekend. But, that's sort of a "Having looked at this many times before, this is my guess" calculation.

What would give you that idea?

Thanks for the information. This helps a lot.

 

[Priceless]

Re: Pairwise and Bracketology 2013 Edition

Is this a a given regardless of what happens over the next 2 weekends?

As far as I can tell. They will win COp and have a big enough lead in TUC that they would win that by .040 points even in a worst-case scenario. I don't see any way for Minnesota to catch them.

 

[Steve_MN]

Re: Pairwise and Bracketology 2013 Edition

From reading what the experts have to say on here, you will most likely have to wait until after this weekend to get those percentages, too many variables at this point.

LT is correct. And, this is why:
Remember the 4 pieces of each comparison. RPI, TUC, ComOpp, H2H. At this point, there is likely to be only slight movement in the CommOpp and H2H part of any particular comparison.

However, a team's RPI can move perhaps +/- .100 this weekend, and, in the case of ECAC teams, the same next weekend if they get that far. That has 2 effects: It affects how their RPI stacks up against other teams, and it affects which teams are TUCs.

There are a few schools whose TUC status apparently makes a big difference in the at-large field. Among these are
Michigan (because of Notre Dame)
Merrimack (because of BU)
Colorado College (because of Wisconsin)

That being said, Priceless has remarked that it seems that
Quinnipiac and Minnesota are at this point locks for the field.

I assume the poster from 2 posts up is a Union backer. Without doing a full up recreation of this weekend, I would say that Union has to win this weekend, or it's hard for them, because Dartmouth would likely pass them in the PWR. And, it could be that they also will have to not lose 2 games next weekend. But, that's sort of a "Having looked at this many times before, this is my guess" calculation.

Quinnipiac is the overall #1 seed, will be #1 in Providence.
Minnesota has clinched, can only be knocked from a 1-seed by a weird confluence of events, and will likely be the 1-seed in Grand Rapids.
Miami has clinched and will likely be the 1-seed in Toledo.

Everyone else still has work to do. UND and Denver can clinch a spot by winning this weekend. Lowell, Boston College and New Hampshire would clinch with wins. New Hampshire is very close but plays TUC Providence this weekend on the road which really hurts them if they are swept. I don't think anyone else from the CCHA or ECAC can clinch this weekend.

I went back to look at last year and only Boston College had clinched a spot before the conference quarters. A bunch of teams were able to clinch by winning those series. In fact, the field was pretty much set last year, barring what we referred to as "the parley" with a bunch of upset winners in the conferences (Bowling Green, Providence, Harvard etc). What was still to be determined was the final order of the teams.

I also found this nugget, which came to mind when the DIY calculator put UNH 4th and UND 5th in an earlier computation:

According to:

http://www.playoffstatus.com/ncaahockey/ncaahockeytournseedprob.html

QU and MN are the only true "locks" for the tournament, with QU being locked in to a top 4 seed (almost certainly the #1 overall) with MN effectively being locked in to one of the remaining #1 seeds (almost certainly the #2 overall) with somewhere less than 1% of dropping below the #4 overall, but no lower than 11th.

Now, I'm not going to say I fully buy in to how they calculate the winning percentages (they use a "Meaningful Win Percentage" number they've come up with) due to the fact that I have zero idea how this is calculated. Having said that, I have to assume that they use that the way we might use KRACH to figure out the win probabilities for any given game, then weighting that in a big spreadsheet of all the games to figure out how things could drop. Assuming that this is anywhere close to accurate, I can't imagine that they'd post a true 0% chance of missing the field unless there was literally no combination of possible results that would result in that happening, so I do assume their numbers are accurate in that respect, at least.

Assuming that they run the calculations even vaguely close to my description, it looks like the following teams are also effectively locks (a less than 1% chance of missing the tourney, which amounts to some bizarre combination of results kicking them out):
Miami, UML, Mankato, North Dakota, Notre Dame
With BC, Denver, UNH and WMU all being in the 1-3% chance of missing. After that, it's all double-digits.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

According to:

http://www.playoffstatus.com/ncaahockey/ncaahockeytournseedprob.html

QU and MN are the only true "locks" for the tournament, with QU being locked in to a top 4 seed (almost certainly the #1 overall) with MN effectively being locked in to one of the remaining #1 seeds (almost certainly the #2 overall) with somewhere less than 1% of dropping below the #4 overall, but no lower than 11th.

Now, I'm not going to say I fully buy in to how they calculate the winning percentages (they use a "Meaningful Win Percentage" number they've come up with) due to the fact that I have zero idea how this is calculated. Having said that, I have to assume that they use that the way we might use KRACH to figure out the win probabilities for any given game, then weighting that in a big spreadsheet of all the games to figure out how things could drop. Assuming that this is anywhere close to accurate, I can't imagine that they'd post a true 0% chance of missing the field unless there was literally no combination of possible results that would result in that happening, so I do assume their numbers are accurate in that respect, at least.

Assuming that they run the calculations even vaguely close to my description, it looks like the following teams are also effectively locks (a less than 1% chance of missing the tourney, which amounts to some bizarre combination of results kicking them out):
Miami, UML, Mankato, North Dakota, Notre Dame
With BC, Denver, UNH and WMU all being in the 1-3% chance of missing. After that, it's all double-digits.

Thanks for that link. It sounds familiar, like we talked about it last year, too. I sent them an email just now, asking for detail in how they do the calculation. More specifically, I asked if they run all the possibilities, and then compute PWR at the end of each, or if they have some other method.

Given that they have Mankato always ranked ahead of NoDak, etc, I have a funny guess that they are actually running a simulation of games, and then calling the field "Tourney Winners plus the Next X number of teams to make 16 altogether in our ranking system." In which case, since they are actually not doing a PWR, it wouldn't be as accurate as it could be.

But, we shall see.

 

[Fighting Sioux 23]

Re: Pairwise and Bracketology 2013 Edition

Thanks for that link. It sounds familiar, like we talked about it last year, too. I sent them an email just now, asking for detail in how they do the calculation. More specifically, I asked if they run all the possibilities, and then compute PWR at the end of each, or if they have some other method.

Given that they have Mankato always ranked ahead of NoDak, etc, I have a funny guess that they are actually running a simulation of games, and then calling the field "Tourney Winners plus the Next X number of teams to make 16 altogether in our ranking system." In which case, since they are actually not doing a PWR, it wouldn't be as accurate as it could be.

But, we shall see.

Keep in mind that they use "Meaningful Winning Percentage" (their ranking system) to determine those "seedings." For example, they currently have Niagara at #16, when they are t13 in the PWR. They also have Providence at #15...they are t21 in the PWR. They have Mankato at #5, when in the PWR they are #9. They calculate things using their own system...it has little to do with how the teams are actually selected. It would be akin to simulating games based on KRACH, and then seeding teams based on their KRACH scores after those simulations. It's simply irrelevant.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Keep in mind that they use "Meaningful Winning Percentage" (their ranking system) to determine those "seedings." For example, they currently have Niagara at #16, when they are t13 in the PWR. They also have Providence at #15...they are t21 in the PWR. They have Mankato at #5, when in the PWR they are #9. They calculate things using their own system...it has little to do with how the teams are actually selected. It would be akin to simulating games based on KRACH, and then seeding teams based on their KRACH scores after those simulations. It's simply irrelevant.

Thanks FS.

That was the way I figured it, too. And, I think we went over this last year. I wanted to make sure, so I sent them an email earlier today. No reply yet.

 

[Patman]

lowell doesnt move up , not playing a tuc , Had them sweeping maine .

Uhm, RPI goes up as well... Of course I haven't played w any of the calculators

 

[Patman]

Thanks FS.

That was the way I figured it, too. And, I think we went over this last year. I wanted to make sure, so I sent them an email earlier today. No reply yet.

again all the more reason to get a good system in place, but its always complicated

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

again all the more reason to get a good system in place, but its always complicated

Not sure I understand this comment in context Patman. Are you saying the current system is or is not good? are you commenting on the site at playoffstatus.com? Do you know those people? Sorry. I am slow. I don't understand....

It is my assumption from looking at that site, and from what FS wrote, that they have their own system, and they are not aware that it doesn't match the PWR. And, that they don't know how the PWR works. Do you know more about that?

 

[Steve_MN]

Re: Pairwise and Bracketology 2013 Edition

Keep in mind that they use "Meaningful Winning Percentage" (their ranking system) to determine those "seedings." For example, they currently have Niagara at #16, when they are t13 in the PWR. They also have Providence at #15...they are t21 in the PWR. They have Mankato at #5, when in the PWR they are #9. They calculate things using their own system...it has little to do with how the teams are actually selected. It would be akin to simulating games based on KRACH, and then seeding teams based on their KRACH scores after those simulations. It's simply irrelevant.

This is what I'm not sure about. IF they're using their MWP ranking system to sort through the games, but then running them through Pairwise at the end, then this has SOME utility (at least in showing what possible outcomes are there). If they're just taking the top 15 in their own rating after going through the "simulation" then yes, this is completely useless.