Pairwise and Bracketology 2013 Edition

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

OK. Full-up example of how the committee process works.
First, the top 16, separated into seeding bands (Note that in this case, we assume all conference champions are part of the top 16.)

#1s: QU, Minny, Miami, Lowell
#2s: ND, BC, Mankato, Denver
#3s: Yale, UNH, Niagara, SCSU
#4s: WMU, NoDame, Union, Wisco

The Committee next has instructions to place the #1s, in order from 1 - 4, in the regional closest to home, that is still available:
QU (1 overall) to Providence
Minny to Rapids
Miami to Toledo
Lowell to Manchester

#2s:
Denver (#8 overall) to Providence with QU
Mankato (7) to Rapids with Minny
BC (6) to Toledo with Miami (Note that even if ND had been 6 and BC 5, BC still goes to Toledo to avoid UNH)
NoDak (5) to Manchester

#3s: Note that SCSU has to pair with BC, because all the other #2s are WCHA clubs
Yale (9) to Providence
UNH (10) to Manchester (This is out of order, but they are host at Manchester, and thus are required to go there)
Niagara (11) to Rapids
SCSU (12) to Toledo (To face BC, as noted above.)

#4s: First observe that neither WMU nor NoDame can be in Toledo, and face Miami, so
Wisco (16) to Providence
Union (15) to Toledo so that Miami doesn't match WMU or NoDame
NoDame (14) to Rapids
WMU (13) to Manchester.

Thus:
Providence: QU(1), Denver (8), Yale (9), Wisco (16)
Rapids: Minny (2), Mankato (7), Niagara (11), NoDame (14)
Toledo: Miami (3), BC (6), SCSU (12), Union (15)
Manchester: Lowell (4), NoDak (5), UNH (10), WMU (13)

Only real option for changes I see here would be to swap Mankato/Niagara game for BC/SCSU to get Niagara to Toledo, which saves a flight cost and might help Toledo attendance.

This explanation is done, not as prediction, but to illustrate the process for those who are new here.

 

[Fishhawk]

Re: Pairwise and Bracketology 2013 Edition

Did Merrimack drop from being a TUC tonight?

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

A few comments on moves in the PWR from tonight's games:

Because their overall positions were solid, neither QU nor Miami suffered from losing.
BU defeating Merrimack caused Merrimack to drop out of the ranks of TUCs, and thus, winning the game cost BU a few comparisions.
Notre Dame PWR rank continues to be several points below their RPI rank.
Wisconsin was helped by Denver tonight, because CC dropped from being a TUC. Again, the effect of this was to bring Wisco's TUC record in line with their RPI. A sort of normalization.

For tomorrow night:
CCHA:
If Michigan again defeats WMU, this will change, as Michigan will become a TUC, and NoDame was 4-0 vs Michigan.
In contrast, if Ferris again beats OSU, then No Dame will lose a 1-0-1 from their TUC record. However, I believe that if both of these happen, then Notre Dame will still rise in the PWR in comparison to their RPI.
About OSU: Miami was 3-1-1 against them. So, if FSU beats OSU again, then Miami's TUC record drops to 9-5-1. This is still very good, and would not change anything. Add their 1-1 with Michigan, and still nothing changes.
WCHA:
CC winning the next 2 would hurt Wisconsin for the above reason in reverse.
Miami and QU are still more (QU) or less (Miami) safe.
The Bubble (in this case from #15 Union down to maybe Cornell at #21, is very unpredictable due to close RPIs.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Did Merrimack drop from being a TUC tonight?

Yes, and see #563 for more details on other things related as well.

 

[C-H-C]

Re: Pairwise and Bracketology 2013 Edition

If PRovidence beats UNH Saturday or Sunday and goes on to beat either BC or Lowell in the Hockey East semifinals, would PC move ahead of UNH and earn an NCAA invitation? Would that alone knock UNH out of the NCAA's?

If PC moves ahead of UNH and Boston University wins the Hockey East Championship, would that knock UNH out (BU earning the autobid)?

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

If PRovidence beats UNH Saturday or Sunday and goes on to beat either BC or Lowell in the Hockey East semifinals, would PC move ahead of UNH and earn an NCAA invitation? Would that alone knock UNH out of the NCAA's?

If PC moves ahead of UNH and Boston University wins the Hockey East Championship, would that knock UNH out (BU earning the autobid)?

Quick answer: Because UNH has such a high RPI compared to the bubble (which I am calling #15 and down, due to the very large RPI difference between Union, at 15th, and the next highest team), they are pretty well insulated from falling out even with a loss tomorrow night. In the event that all the home teams win again tomorrow, it seems that from HE: Lowell, BC, UNH(in spite of losing) will all be in. Providence will need to win the tourney (at least under likely scenarios), and BU is more likely to need win the tourney than to find their way in as an at-large.

 

[Patman]

Re: Pairwise and Bracketology 2013 Edition

Question, what's the minimum number of games of Division I games needed for a team to qualify for their games to be considered for tournament qualification (Penn State, in this instance?). I am trying to re-create the RPI and i'm having challenges getting what others have... I'm trying to figure out where this error is occurring.

edit: nevermind... I was looking at the wrong column (using jtw's version)... OppOpp still doesn't match... I'll take tonight's results as soon as jtw provides an update and match against Sioux Sports.

edit #2: ok, no problems at all... I just suck at looking at columns

 

[Heavy]

Re: Pairwise and Bracketology 2013 Edition

Hey everyone, I'm new to the thread and have some general questions about calculating these ratings. First, does anyone do these calcs on their own? I would love to mess with a script or tool that allowed me to play with all the teams at once. I have written a couple Bradley-Terry solutions using Solver in Excel and a system of equations in Mathematica but I haven't done it for all the teams. It gets kind of unwieldy once I start trying to handle more than about 10 teams. I did the WCHA last year just for the fun of it, but of course it doesn't mean much to omit all the other teams.

I have also looked at the build your own rankings site. That's pretty cool but it's kind of a black box. Does anyone know who wrote the cgi script and if we could open it up a bit? I'd be willing to host a site that runs the script but I'd like to see more of the inner workings, as that's honestly what's most interesting to me. If we had a tool that was easy to use and allowed people to see a bit more of the magic, I think that would be of interest to a lot of us more technical hockey fans.

Any thoughts are appreciated.

 

[Patman]

Re: Pairwise and Bracketology 2013 Edition

Hey everyone, I'm new to the thread and have some general questions about calculating these ratings. First, does anyone do these calcs on their own? I would love to mess with a script or tool that allowed me to play with all the teams at once. I have written a couple Bradley-Terry solutions using Solver in Excel and a system of equations in Mathematica but I haven't done it for all the teams. It gets kind of unwieldy once I start trying to handle more than about 10 teams. I did the WCHA last year just for the fun of it, but of course it doesn't mean much to omit all the other teams.

I have also looked at the build your own rankings site. That's pretty cool but it's kind of a black box. Does anyone know who wrote the cgi script and if we could open it up a bit? I'd be willing to host a site that runs the script but I'd like to see more of the inner workings, as that's honestly what's most interesting to me. If we had a tool that was easy to use and allowed people to see a bit more of the magic, I think that would be of interest to a lot of us more technical hockey fans.

Any thoughts are appreciated.

Dig down for the eLynah link to jtw's (john whelan's) rankings... its a bit modifiable. I'm working on my own stuff but it sure as hell isn't going to be a web-script. I'd like to think I'm building (again) to something bigger down the line.

In the end, EXCEL isn't going to cut it that easily. I mean, you could, but that'd be a pain in the ***.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Hey everyone, I'm new to the thread and have some general questions about calculating these ratings. First, does anyone do these calcs on their own? I would love to mess with a script or tool that allowed me to play with all the teams at once. I have written a couple Bradley-Terry solutions using Solver in Excel and a system of equations in Mathematica but I haven't done it for all the teams. It gets kind of unwieldy once I start trying to handle more than about 10 teams. I did the WCHA last year just for the fun of it, but of course it doesn't mean much to omit all the other teams.

I have also looked at the build your own rankings site. That's pretty cool but it's kind of a black box. Does anyone know who wrote the cgi script and if we could open it up a bit? I'd be willing to host a site that runs the script but I'd like to see more of the inner workings, as that's honestly what's most interesting to me. If we had a tool that was easy to use and allowed people to see a bit more of the magic, I think that would be of interest to a lot of us more technical hockey fans.

Any thoughts are appreciated.

Heavy,
No programmer here. I have done a Bradley-Terry mock up for NHL (30 teams, but it all works the same, of course) in Excel, I mean just the plain spreadsheet. I chose to do it to where I had to do the cycling manually, though. And, I didn't use programming to extract the matrix, I hard inputted it.

So, with what little experience I have, I would say some thing like that for the PWR look daunting to me. One poster here calls it "Simple Math." And, it is, but it isn't.

If you are reading the Build Your Own Rankings on elynah, I am wondering just what else you would want available?

I love that tool. It's all you need at this time of year, to check on many many possibilities. That, along with a couple of other sites (like siouxsports, which gives a run down of CommOpp and TUC), is enough to see which compares can change.

The only thing I think is lacking from elynah is a way to grade the TUC results. Further up here, Patman and I discussed ways to do that - to make a result vs QU worth more than vs Brown. That way, TUC is a slope instead of a cliff.

So, again, what more are you looking for?

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

Question, what's the minimum number of games of Division I games needed for a team to qualify for their games to be considered for tournament qualification (Penn State, in this instance?). I am trying to re-create the RPI and i'm having challenges getting what others have... I'm trying to figure out where this error is occurring.

edit: nevermind... I was looking at the wrong column (using jtw's version)... OppOpp still doesn't match... I'll take tonight's results as soon as jtw provides an update and match against Sioux Sports.

edit #2: ok, no problems at all... I just suck at looking at columns

So, Patman, you are good to go? i have the actual script for tonight's games to input to elynah if you need that.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

I am remembering Priceless' list of teams who are out; who need to win their conference tourney; who can still be at-large; and who are in.

Does anyone know of a way for Omaha to get in now without winning the WCHA?

 

[Heavy]

Re: Pairwise and Bracketology 2013 Edition

The Excel piece is only of interest because everyone has it and is comfortable using it. I'd prefer to do something in Python or perhaps R. I'm not really talking about splitting the atom, just a way to input records easily and calculate kratch, etc. Adding user defined extras could come later.

Really, the Build Your Own Rankings is most of what I'd like to do. I'd just like to see the source and make it available to others. Does anyone know this Joe Schlobotnik fella? Maybe I'll contact him.

 

[Patman]

Re: Pairwise and Bracketology 2013 Edition

So, Patman, you are good to go? i have the actual script for tonight's games to input to elynah if you need that.

I'm good to go... to sleep... to count sheep. Tomorrow, I'll get to it tomorrow. I should be able to wrap the rest up... btw, I don't promise any simulators by the end of the weekend, but I can't rule it out either.

---
Heavy, I'm working on R code but I may not necessarily want to float it out there. We'll see how generous I feel. One thing's for sure, I feel this is a lot more efficient than the last time I did this. Work on two major national level projects and you tend to sharpen your skills

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

And, since I am in a particularly windy mood tonight. Notice that right now, there are 6 WCHA teams in the field. Some fans may wonder if that can continue. I believe the answer is yes:
Minny is safe
I believe that Mankato, NoDak and Denver are also safe
SCSU should be safe unless they lose twice to UAA
That leaves Wisco. To keep Wisco in the field: Denver beats CC again tomorrow (keeps CC out of TUC), Wisco beats UMD again, and then I think at the Final 5, all they need is to win on Thursday, which would likely be against Mankato or Omaha.

Not too complicated, really.

 

[Heavy]

Re: Pairwise and Bracketology 2013 Edition

If you are reading the Build Your Own Rankings on elynah, I am wondering just what else you would want available?

I love that tool. It's all you need at this time of year, to check on many many possibilities. That, along with a couple of other sites (like siouxsports, which gives a run down of CommOpp and TUC), is enough to see which compares can change.

Those tools are super cool. I'm just the kind of person that wants to run them on my own, mostly for fun and to learn a thing or two, but also to make them available to others.

 

[Almington]

Re: Pairwise and Bracketology 2013 Edition

And, since I am in a particularly windy mood tonight. Notice that right now, there are 6 WCHA teams in the field. Some fans may wonder if that can continue. I believe the answer is yes:
Minny is safe
I believe that Mankato, NoDak and Denver are also safe
SCSU should be safe unless they lose twice to UAA
That leaves Wisco. To keep Wisco in the field: Denver beats CC again tomorrow (keeps CC out of TUC), Wisco beats UMD again, and then I think at the Final 5, all they need is to win on Thursday, which would likely be against Mankato or Omaha.

Not too complicated, really.

If UNO loses to Mankato tomorrow does UNO drop frpm being a TUC?

If all the higher seeds win, UW (4) would play DU (5) in the Thursday afternoon game, UND (3) would play MSUM (6) in the Thursday evening game

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

If UNO loses to Mankato tomorrow does UNO drop frpm being a TUC?

If all the higher seeds win, UW (4) would play DU (5) in the Thursday afternoon game, UND (3) would play MSUM (6) in the Thursday evening game

Omaha would fall from being TUC (.4988). Sorry about the mistake on the FF seedings. Same reasoning applies. UW goes win tonight, DU beats CC again, UW beats DU and I think UW would make the field.

Under this scenario, without accounting for other conferences, I get UW at 16th in the PWR. Right on the bubble.

 

[Patman]

Re: Pairwise and Bracketology 2013 Edition

RPI game removal question... do you remove the game for ALL teams and re-compute, or do you only adjust the team that gets affected?

edit: this may answer why I got bad results last time I tried to do this... my understanding was that you delete and re-start.

 

[Numbers]

Re: Pairwise and Bracketology 2013 Edition

RPI game removal question... do you remove the game for ALL teams and re-compute, or do you only adjust the team that gets affected?

edit: this may answer why I got bad results last time I tried to do this... my understanding was that you delete and re-start.

Patman,
Do you mean the "Game result removed do to negative effect in defeating a poor team."? If so, I believe the explanation I have seen is like this: RPI can also be seen as a sum of results for each game. In other words RPI=sumoverallgames(result of game(w/l/t)*25%+oppforthatgamewin%*21%)+oppoppwin%forthatgame*54%.

Looked at it that way, it becomes obvious that you remove that game only when calculating that teams' RPI.

Another example, In the Women's this year, Minnesota is undefeated. That means that any results against any teams except the strongest team on their schedule will only drop their RPI. So, in their RPI, only the games against NoDak are counted. Obviously, however, those games count in everyone else's RPI.