Pairwise and Bracketology 2013 Edition

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

A question for JimDahl in regard to UNH.
First some background: Save for 4 scenarios (out of ~300,000), UNH is either a 2 or a 3. Now, the question that I have is:

If UNH is 8th overall, that is Q's normal bracket, but Q is closer to Prov. So, would the committee put Q in Prov anyway, and then the 4th seed ends up with the 8th seed? Or, would they put Q in Manchester for bracket integrity? Now, what if UNH is 7th? It's a little easier for the committee, because now if you put Q with the #8, you get 1/8, 4/7, 2/6, 3/5. And, if UNH is 6th, there is no question what they would do.
Likewise, what about if UNH is 9th overall? In that case, I think it's pretty clear Q goes to Prov, and the bracket goes 1/8/10/16; 2/7/11/15; 3/6/12/14; 4/5/9/13 or something. I don't think the committee would concern themselves with UNH as a 3.

So, the question for Jim:
What are the odds for UNH, not by bracket seed, but by overall seed?

 

[Steve_MN]

Re: Pairwise and Bracketology 2013 Edition

That's interesting. Personally I do think they would put QU in Manchester here... not much difference between playing the #15 in, say, Toledo, than playing the 13 in Manchester. I'm sure Quinnipiac would WAY rather be in Manchester.

This 'hosting' nonsense is ****ing stupid.

Host schools are placed in their own region for every sport except D1 Men's Hoops.... learn to live with it.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Or, PatMan, did you save your data from your 10,000 runs? You had it by bracket seed (1,2,3,4). What about by overall seed? I think only UNH is interesting in this right now.

The only other case where it enters is the question:
If Brown wins ECAC, are they always #16? And, I think the answer to that would be: No. AHA teams, ColoColl, and Mich would all come in below them with upset runs.

 

[unh1982]

Re: Pairwise and Bracketology 2013 Edition

•Atlantic Hockey Semifinal #1: Niagara def. Canisius
•Atlantic Hockey Semifinal #2: Connecticut def. Mercyhurst
•Atlantic Hockey Championship: Niagara def. Connecticut
•CCHA Semifinal #1: Miami def. Michigan
•CCHA Semifinal #2: Notre Dame def. Ohio State
•CCHA Championship: Miami def. Notre Dame
•ECAC Semifinal #1: Quinnipiac def. Brown
•ECAC Semifinal #2: Yale def. Union
•ECAC Championship: Yale def. Quinnipiac
•ECAC Consolation: Brown def. Union
•Hockey East Semifinal #1: Providence def. Mass.-Lowell
•Hockey East Semifinal #2: Boston University def. Boston College
•Hockey East Championship: Providence def. Boston University
•WCHA Semifinal #1: St. Cloud State def. Wisconsin
•WCHA Semifinal #2: Minnesota def. North Dakota
•WCHA Championship: Minnesota def. St. Cloud State
•WCHA Play-in #1: North Dakota def. Colorado College
•WCHA Play-in #2: Wisconsin def. Minnesota ..................................I got both bu and pc in , took a while ..cool... odds have no idea

 

[Patman]

Or, PatMan, did you save your data from your 10,000 runs? You had it by bracket seed (1,2,3,4). What about by overall seed? I think only UNH is interesting in this right now.

The only other case where it enters is the question:
If Brown wins ECAC, are they always #16? And, I think the answer to that would be: No. AHA teams, ColoColl, and Mich would all come in below them with upset runs.

I saved nothing... I didn't want a substantially large object holding up the analysis... To be honest, I could probably handle it in terms of RAM.

Fact is, if I set out all results, yes, I could re-produce later, but that can lead to some management issues.

For my sake, a lot of this is still developmental... Unfortunately I am not set up to have easily accessible cross-tabs and all that.

---

I suppose all the more reason to throw open the code.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

I saved nothing... I didn't want a substantially large object holding up the analysis... To be honest, I could probably handle it in terms of RAM.

Fact is, if I set out all results, yes, I could re-produce later, but that can lead to some management issues.

For my sake, a lot of this is still developmental... Unfortunately I am not set up to have easily accessible cross-tabs and all that.

---

I suppose all the more reason to throw open the code.

I see JimDahl has exactly what I want on his blog. Thanks PatMan, this is one of my favorite weeks of the year, and I like connecting with you and Priceless through the process.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Another Shout Out to JimDahl,

I was looking at NoDame's possibilities on your blog. I notice they have a wide range, and a flat distribution compared to others. What interests me is the bump at 10, 11 if they win out, and correspondingly at 15,16 if they lose out. Do you have any idea what other games are responsible for that?

Thanks.

 

[Patman]

I see JimDahl has exactly what I want on his blog. Thanks PatMan, this is one of my favorite weeks of the year, and I like connecting with you and Priceless through the process.

As a math guy I find this stuff interesting by default... I'm interested in evaluating chance and prediction. I am only limited by the ability to process via code. But I suppose this is why great coders get to solve the problems, they have the means.

Analogously, I've found it fascinating that many social and other organizations are run by IT types because they are the ones who know how to set up websites, messaging, scripts, etc.

 

[Craig P.]

Re: Pairwise and Bracketology 2013 Edition

Another Shout Out to JimDahl,

I was looking at NoDame's possibilities on your blog. I notice they have a wide range, and a flat distribution compared to others. What interests me is the bump at 10, 11 if they win out, and correspondingly at 15,16 if they lose out. Do you have any idea what other games are responsible for that?

Thanks.

I know that our position is very TUC-sensitive at the moment, and both Michigan (positively) and Ohio State (negatively) are hanging around the TUCliff (tm). I would hypothesize that these are two of the drivers of the break points. In particular, I think ND's lose-out 10/11 vs. 15/16 division mostly comes down to whether Michigan stays a TUC and whether Ohio State wins out.

 

[goblue78]

Re: Pairwise and Bracketology 2013 Edition

Yet one more question to JimDahl (combined with a big thanks). I did some calculations from your blog posting and I get that there are around 4500 scenarios in which Yale doesn't get in but didn't lose two games. I assume that virtually all of those are a loss and a tie. Is there any way for Yale to go 1-1-0 and not be safe?

 

[King]

Re: Pairwise and Bracketology 2013 Edition

Here is a fairly reasonable example for BU:

Atlantic Hockey
Semifinal #2: Connecticut defeats Mercyhurst
Semifinal #1: Niagara defeats Canisius
Championship game: Niagara defeats Connecticut
CCHA
Semifinal #2: Notre Dame defeats Ohio State
Semifinal #1: Miami defeats Michigan
Championship game: Miami defeats Notre Dame
ECAC
Semifinal #2: Yale defeats Union
Semifinal #1: Quinnipiac defeats Brown
Championship game: Quinnipiac defeats Yale
Consolation game: Brown defeats Union
Hockey East
Semifinal #2: Boston University defeats Boston College
Semifinal #1: Massachusetts-Lowell defeats Providence
Championship game: Massachusetts-Lowell defeats Boston University
WCHA
Play-in #2: North Dakota defeats Colorado College
Play-in #1: Minnesota State defeats Wisconsin
Semifinal #2: Minnesota defeats North Dakota
Semifinal #1: St. Cloud State defeats Minnesota State
Championship game: Minnesota defeats St. Cloud State


Read more: http://www.uscho.com/rankings/pairwise-predictor/#ixzz2NuoR7APL

The key here seems to be Union losing twice.

Actually, BU gets in as #16 if you run that same scenario and Union wins the consolation game (actually, I had Wisconsin winning the play-in and losing semi to presumably make it tougher for BU).

If an "outsider" (PC, CC, UConn etc) wins a tourney to grab the 16th spot, then BU would need Union to lose twice to finish 15th.

 

[ExileOnDaytonStreet]

Re: Pairwise and Bracketology 2013 Edition

I don't like UW's chances if between Union, BU, Providence and Ohio State there aren't three of them that kind of wet the bed in the semis.

Niagara winning the AHA is evidently a pretty big deal for us, even aside from the "AHA only gets one spot instead of two" angle.

 

[Umileated]

Re: Pairwise and Bracketology 2013 Edition

I see JimDahl has exactly what I want on his blog. Thanks PatMan, this is one of my favorite weeks of the year, and I like connecting with you and Priceless through the process.

Indeed. I've been mostly just reading the last couple of weeks, but it definitely makes me wish I had chosen a concentration in stats.

 

[Alaska88]

could be wrong but i see both pc and bu needing autobids , henceforth there should be just two on the bubble as you have described.

What is the scenario for Mercyhurst to become a TUC?

 

[JimDahl]

Re: Pairwise and Bracketology 2013 Edition

Yet one more question to JimDahl (combined with a big thanks). I did some calculations from your blog posting and I get that there are around 4500 scenarios in which Yale doesn't get in but didn't lose two games. I assume that virtually all of those are a loss and a tie. Is there any way for Yale to go 1-1-0 and not be safe?

Sorry, didn't make it back over here this afternoon. As you discovered, I now have full probabilities by team/by ranking/by number of wins posted:

Remaining PWR possibilities

As to your question, nope I don't see any scenarios where Yale gets 1 win, 1 loss, and doesn't make the tournament.

What is the scenario for Mercyhurst to become a TUC?

That one is tricky, getting their RPI to .500 requires everything to go just so. Here's an example:


Atlantic Hockey Semifinal #1: Niagara def. Canisius
Atlantic Hockey Semifinal #2: Mercyhurst def. Connecticut
Atlantic Hockey Championship: Mercyhurst def. Niagara
CCHA Semifinal #1: Michigan def. Miami
CCHA Semifinal #2: Ohio State def. Notre Dame
CCHA Championship: Michigan def. Ohio State
ECAC Semifinal #1: Brown def. Quinnipiac
ECAC Semifinal #2: Union def. Yale
ECAC Championship: Brown def. Union
ECAC Consolation: Quinnipiac def. Yale
Hockey East Semifinal #1: Mass.-Lowell def. Providence
Hockey East Semifinal #2: Boston University def. Boston College
Hockey East Championship: Mass.-Lowell def. Boston University
WCHA Semifinal #1: Minnesota State def. St. Cloud State
WCHA Semifinal #2: North Dakota def. Minnesota
WCHA Championship: Minnesota State def. North Dakota
WCHA Play-in #1: North Dakota def. Colorado College
WCHA Play-in #2: Minnesota State def. Wisconsin

 

[goblue78]

Re: Pairwise and Bracketology 2013 Edition

Thanks again, Jim. The extra column is really helpful. The ECAC teams are still missing the "One loss and one tie" columns, but we probably deserve it for being the last conference to keep the consolation game.

 

[fairbanks hockey puck]

Re: Pairwise and Bracketology 2013 Edition

Got AK to 16 alone
-----------------------------
Atlantic Hockey

Semifinal #2: Connecticut defeats Mercyhurst
Semifinal #1: Niagara defeats Canisius
Championship game: Niagara defeats Connecticut

CCHA

Semifinal #2: Notre Dame defeats Ohio State
Semifinal #1: Miami defeats Michigan
Championship game: Notre Dame defeats Miami

ECAC

Semifinal #2: Union defeats Yale
Semifinal #1: Quinnipiac defeats Brown
Championship game: Union defeats Quinnipiac
Consolation game: Yale defeats Brown

Hockey East

Semifinal #2: Boston College defeats Boston University
Semifinal #1: Providence defeats Massachusetts-Lowell
Championship game: Boston College defeats Providence

WCHA

Play-in #2: North Dakota defeats Colorado College
Play-in #1: Wisconsin defeats Minnesota State
Semifinal #2: Minnesota defeats North Dakota
Semifinal #1: St. Cloud State defeats Wisconsin
Championship game: Minnesota defeats St. Cloud State

 

[fairbanks hockey puck]

Re: Pairwise and Bracketology 2013 Edition

a lot of if's, but's and or's

 

[Priceless]

Re: Pairwise and Bracketology 2013 Edition

This year's version of the parley is

Connecticut or Mercyhurst defeat Niagara in the AHA
Brown wins the ECAC
Michigan or Ohio State wins the CCHA
Colorado College wins the WCHA
Providence or Boston University win HockeyEast

If all those things happen the cut line

Weird stuff starts happening, like the #1 seed in the WCHA tournament losing the championship game and missing the NCAA tournament. Teams on the bubble will feel much more secure if North Dakota takes care of business on Thursday and eliminates CC. It would also help if Mankato ended Bucky's season.

Last year 12 teams had clinched (plus the AHA champ) and 6 teams were going for the last 3 spots. If things break right this season, 14 of the 16 teams could already have clinched by Saturday morning.

 

[goblue78]

Re: Pairwise and Bracketology 2013 Edition

In an attempt to do some new analysis to add to the great stuff JimDahl, Priceless and Numbers have done, I have calculated the KRACH-based distribution of cut lines. This is completely accurate except that it ignores the possibility that Providence can, if it wins HE, qualify anyway, which is a really small possibility and is swamped by the imprecision of KRACH.
Cut Line
16 0.168953791
15 0.404773548
14 0.318417224
13 0.097783858
12 0.009818925
11 0.000252654

So Priceless' nightmare scenario, a cut line of 11, isn't very likely (under Krach). Together, a cut line of 14 or 15 has a probability of almost 75%. And the never-before-seen cutline of 16, the pure PWR scenario, has a probability of 17 percent.