Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

[TonyTheTiger20]

It took several hours of fixing last year's spreadsheet thanks to Merrimack's rise to D1, but after opening up the laptop on a couple trans-Pacific flights in the offseason I was able to update the Pairwise Predictor for the 2015-2016 season.

You can play with results here and see how it would affect the Pairwise (i.e. if St. Lawrence wins tomorrow, if Minnesota had lost yesterday, etc.).

Any questions or bug reports, please let me know!

Note that my spreadsheet will disagree with USCHO's PWR/RPI until everyone (other than the Ivies, anyway) has played two opponents and every opponent has played two opponent's opponents. If a team is listed on my spreadsheet, it means they fit that criteria. If they have an RPI of 0.000, it means that they either haven't started playing yet (Ivies) or someone they've played hasn't played another opponent. The reason is that "Opponents Winning Percentage" is a divide-by-zero if your opponent hasn't played anyone else. USCHO appears to change divide-by-zero errors into zeroes to make it not break the formula, I just change the team's entire RPI to zero since they're both useless corrections anyway.

Short version: In a week or two all the non-Ivies should match up to USCHO's PWR/RPI, and until then the ones on my spreadsheet are accurate with no divide-by-zero errors.

[KTDC]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Timothy A's comment about the Gopher's not trying to score in the third period was legitimately funny, come on!

But, who has scored more than 8 times in a game since the 2011-12 Season.

Here's some:
Minnesota:
10/10/15...St Cloud...11-0
1/4/15...St Lawrence...10-0
12/5/14...St Cloud....12-0
2/7/14...Bemidji State...10-0
12/7/13...Ohio State...9-2
12/1/13...Princeton...9-1
11/24/12...New Hampshire...10-2
11/17/12...Minnesota State...9-1
9/29/12...Colgate...11-0
11/18/11...New Hampshire...11-0

BC
1/11/15...Northeastern...9-1
11/28/14...Harvard...10-2
11/15/14...New Hampshire...10-2
10/4/14...Syracuse...10-2
12/1/12...Maine...10-0

Wisconsin
10/6/12...Lindenwood...9-1
9/25/11...Lindenwood...13-1
9/23/11...Lindenwood...11-0

Harvard
1/4/13...Union...9-0
10/27/12...Princeton...9-1
2/4/12...Princeton...10-1

Clarkson
9/28/13...RIT...12-1
12/2/11...Rensselaer...10-0

BU
2/26/12...New Hampshire...9/1

Minnesota has some experience calling off the dogs. The question is: did BC call off the dogs vs Harvard? They win the prize for strongest opponent.

[NUWH DogHouse]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Timothy A's comment about the Gopher's not trying to score in the third period was legitimately funny, come on!

But, who has scored more than 8 times in a game since the 2011-12 Season.

Here's some:
Minnesota:
10/10/15...St Cloud...11-0
1/4/15...St Lawrence...10-0
12/5/14...St Cloud....12-0
2/7/14...Bemidji State...10-0
12/7/13...Ohio State...9-2
12/1/13...Princeton...9-1
11/24/12...New Hampshire...10-2
11/17/12...Minnesota State...9-1
9/29/12...Colgate...11-0
11/18/11...New Hampshire...11-0

BC
1/11/15...Northeastern...9-1
11/28/14...Harvard...10-2
11/15/14...New Hampshire...10-2
10/4/14...Syracuse...10-2
12/1/12...Maine...10-0

Wisconsin
10/6/12...Lindenwood...9-1
9/25/11...Lindenwood...13-1
9/23/11...Lindenwood...11-0

Harvard
1/4/13...Union...9-0
10/27/12...Princeton...9-1
2/4/12...Princeton...10-1

Clarkson
9/28/13...RIT...12-1
12/2/11...Rensselaer...10-0

BU
2/26/12...New Hampshire...9/1

Minnesota has some experience calling off the dogs. The question is: did BC call off the dogs vs Harvard? They win the prize for strongest opponent.

NU dropped 8 on UNH (a shutout) in the first ever women game on ESPN3 in 2012

Plus in '12-'13 9-1 over UConn, 8-1 over UNH 8-7 against Providence

'13-'14 8-2 v Maine

plus a number of 7 goal games

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

I think he's going more than 8 (I.e. 9+)

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

It looks like there are no more errors in the calculation -- every team has played 2-3 opponents (other than the Ivies)

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

The Pairwise spreadsheet should be fully good to go. All teams are connected by two different opponents so there shouldn't be any more divide by zero errors.

If anyone sees anything wrong, let me know

[bc6696]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Can someone give me some insight on the SOS as it pertains to RPI. For instance: Mercyhurst has the 32nd ranked SOS, but has played NU, QU, and Princeton (RIT too) - seems like it should be better than that??

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Can someone give me some insight on the SOS as it pertains to RPI. For instance: Mercyhurst has the 32nd ranked SOS, but has played NU, QU, and Princeton (RIT too) - seems like it should be better than that??

Strength of schedule at this point is really random because it's so early in the season. Keep in mind too that RPI's strength of schedule factors not just NU's record but also the record of the teams NU has played. Those numbers are really wonky at this point.

Once everything settles down in a couple more months and everything kind of regresses to what it should be you'll have a more accurate strength of schedule ranking.

EDIT: Rutter's website breaks out RPI and you can kind of see what I mean: http://math.bd.psu.edu/faculty/rutte...nt_D1_RPI.html

Mercyhurst's opponents have excellent records but their opponents' opponents' records are awful.

[Eeyore]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Mercyhurst's opponents have excellent records but their opponents' opponents' records are awful.

This will look especially goofy early in the season, because the RPI coefficient for opponents' opponents' winning percentage is, by a significant margin, the largest of the three. It has to be for this category to have much effect on the results, because OOWP converges pretty rapidly to .500 once you have a lot of games played. This early in the season, though, that convergence hasn't really happened yet, and so OOWP can have an outsize effect on the computation.

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

This will look especially goofy early in the season, because the RPI coefficient for opponents' opponents' winning percentage is, by a significant margin, the largest of the three. It has to be for this category to have much effect on the results, because OOWP converges pretty rapidly to .500 once you have a lot of games played. This early in the season, though, that convergence hasn't really happened yet, and so OOWP can have an outsize effect on the computation.

Yep this

[5mn_Major]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Once ECAC teams get some more games under their belt.

Teams like UMD will skyrocket. After this weekend, UMD will have 6 of 10 games against teams that have a good claim to be the best in the country.

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

I just noticed that KRACH is not only missing Merrimack but is also ignoring all games involving them:

http://www.uscho.com/rankings/krach/d-i-women/

For example UNH is listed as 2-9-0 even though they also have a loss to MC.

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

I think ARM and Arafel just did this exercise in a Wednesday Women recently. But here is the current Pairwise top 8:

1) UW
2) BC
3) UM
4) Clarkson
5) NU
6) Bemidji
7) QU
8) Harvard

Take our your future seeing machines... of these 8 teams, which do you think will not make the tournament, and who will they be replaced by? We know that at a minimum one will be kicked out for the CHA champion. Anyone else?

The top 3 are obviously going to be in. As for the rest, I think Clarkson is too talented to drop out. NU is really talented and I'm really high on them, but they still have 3 left against BC and have yet to prove that they won't choke their way out of being a TUC, let alone a tournament team.

BSU only has 2 left against UM and 2 left against UW. Will they be a 3rd WCHA team in?

QU has a tough 2nd half and hasn't faced either Harvard or Clarkson yet.

Harvard... I have a really tough time not seeing them in the field.

For my money, the three teams most likely to not make it are QU, BSU, and Northeastern. I have a hard time seeing anyone who is currently outside the top 8 making enough of a run to get in, particularly NoDak because you've already got 3 WCHA teams in front of them. Four is just not mathematically feasible.

Forced to pick 8, I would go with UM, UW, BC, Clarkson, Harvard, BSU, NU, and CHA (call it Penn State because I'm on that bandwagon so I might as well ride it into the ground), in that order.

[wwhyte]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Forced to pick 8, I would go with UM, UW, BC, Clarkson, Harvard, BSU, NU, and CHA (call it Penn State because I'm on that bandwagon so I might as well ride it into the ground), in that order.

Is it conceivable that WCHA teams get the top two positions? I would have thought just on WINP BC are more likely for 1st or 2nd than to be behind two WCHA teams.

[TonyTheTiger20]

Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Is it conceivable that WCHA teams get the top two positions? I would have thought just on WINP BC are more likely for 1st or 2nd than to be behind two WCHA teams.

Yes but I think BC is probably not going to run the table. Assuming UW and UM come close to a split, a BC with a loss or two would probably be behind the two of them.

It's a theory I could actually test out with the predictor! I might just do that tonight when I get home from work.