What's new
By:
USCHO Fan Forum

This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

  • The USCHO Fan Forum has migrated to a new plaform, xenForo. Most of the function of the forum should work in familiar ways. Please note that you can switch between light and dark modes by clicking on the gear icon in the upper right of the main menu bar. We are hoping that this new platform will prove to be faster and more reliable. Please feel free to explore its features.

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

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

zoofer said:
Ted, in a thread discussing who has the most games with so and so goal differential, I hardly think perspective is sad, but you're entitled.....at TTT does it tongue and cheek! :p
Click to expand...



my God you people are nuts. the guy is posting telling everyone that a team was on top of the world at one time and that over the horizon comes a dip.

does that not sound sad to you?
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

CD, I've seen the light on insulting. St Ignatius is always watching. no more wasting good ones on Midwest settlers. unless in person
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TonyTheTiger20 said:
When was the last time St. Cloud was a TUC??
Click to expand...
SCSU finished third in the WCHA in 2010, and then all of its offense graduated, and it has been seventh or eighth thereafter.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Somethings messed up with the women's PWR and/or RPI. BC's SOS is shown to be about the same as UMD. UMD should have a bone crushing SOS where BC should not be in contention. Based on that weird math, BC's hyped SOS catapults it to top RPI and UW drops to third. Then switch over to PWR...and mysteriously UW's RPI makes a comeback to #1.

Bad numbers almost guaranteed.
 
5mn_Major said:
Somethings messed up with the women's PWR and/or RPI. BC's SOS is shown to be about the same as UMD. UMD should have a bone crushing SOS where BC should not be in contention. Based on that weird math, BC's hyped SOS catapults it to top RPI and UW drops to third. Then switch over to PWR...and mysteriously UW's RPI makes a comeback to #1.

Bad numbers almost guaranteed.
Click to expand...

Yeah the USCHO RPI link is wrong while the USCHO PWR link is right. I'm not sure what it's doing. I think maybe the RPI link isn't removing "bad wins"...

I'll have to compare to my spreadsheet to see what it's doing. I'm the meantime my link has accurate RPI down past the TUCs if you're looking to see everyone's RPI past the TUC line.

I can also add correct RPI SOS to see if something is wrong with USCHO's SOS numbers.

EDIT 1: The RPI page is indeed not removing "bad wins." That's the difference there. I'm not sure why it's not going it, it always used to.

EDIT 2: The RPI SOS is being calculated corrected on USCHO. It's calculated at a ratio of (.24/.70) times opponents' winning percentage plus (.46/.70) times opponents' opponents' winning percentage.

Or less confusingly, SOS = [(OppW%)*(.24/.7)]+[(OppOppW%)*(.46/.7)]

I'll add this to my spreadsheet this afternoon. I actually already did it (that's how I can tell you USCHO is right) but can't upload to the host site until I get home.
 
Last edited:
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Also, KRACH is still ignoring all games against Merrimack so it's inaccurate.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TRIPLE POST!

Oh and if you're interested in this sort of mental exercise -- you may be wondering "why is Wisconsin's SOS so much lower than BC's and yet they are ranked higher in the PWR despite both having undefeated records?"

The reason is "bad wins." When you have an undefeated record, the only games that count in your RPI are your "best" opponent. Every other game is thrown out.

In this case, Wisconsin's best opponent is Bemidji State and BC's is (interestingly) St. Lawrence*.

*this is due to SLU's extremely high opponents' winning percentage.

So, while overall Wisconsin's schedule has been dogcrap, (:D), its best opponent has been better than BC's best opponent. And when you think about it, when two teams are undefeated, that's probably a pretty fair way of ranking them.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TonyTheTiger20 said:
So, while overall Wisconsin's schedule has been dogcrap, (:D), its best opponent has been better than BC's best opponent. And when you think about it, when two teams are undefeated, that's probably a pretty fair way of ranking them.
Click to expand...
It might be fair in certain cases, but there would definitely be cases where it would be unfair, and as with most things, RPI has to rely on dumb luck because it is incapable of making a meaningful calculation. Team A and Team B could be undefeated at this point. Team A could play most of its games against teams near the bottom, and one game against a team just a bit better than Team B's best opponent. Meanwhile, Team B could be playing all of its games against teams of relatively even strength, meaning most of its "bad" games are much better results than those of Team A. Due to its limitations, RPI is forced to discard the bulk of the information available to it, because it lacks an ability to do anything useful with it. Thus, RPI would judge them on the basis of one game apiece, while ignoring the fact that Team A is playing a weaker schedule overall.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

I've been thinking about the strength of schedule question lately. RPI's SOS calculation seems to be extra crappy for some reason and I've been curious about why.

College hockey calculates RPI in a different way than most sports. Most sports take Win%/OppWin%/OppOppWin% as a .25/.50/.25 split.

Women's hockey not only has a different split at .30/.24/.46, but it emphasizes a different value entirely -- instead of putting the most weight on opponent's winning percentage like just about every sport, it puts the most weight on opponent's opponents' winning percentage.

I'm not really sure why. OppOppWin% is the value that converges to 0.500 the quickest and seems borderline random. For example look at Duluth -- it doesn't take a calculator to tell that they have the toughest strength of schedule, and yet their OppOppWin% -- again, the biggest factor in calculating RPI -- is below .500!

Which brings me to strength of schedule. Strength of schedule in RPI is calculated as the ratio of the last two numbers, which as we know is wildly different than in any other sport.

Just for kicks, here is what the RPI would look like if women's hockey used a standard .25/.50/.25 split instead of what is used now:

1) Wisconsin .7985
2) BC .7819
3) Minnesota .6854
4) Bemidji State .6335
5) Clarkson .6243
6) Quinnipiac .6180
7) Northeastern .5872
8) Minnesota-Duluth .5894 !!!!!!!
9) Harvard .5735
10) St. Lawrence .5842
11) Princeton .5720
12) North Dakota .5551

That is insane. Duluth's RPI goes from 18th in the country (.5048), and barely a TUC, all the way up to 8th!!! in the country at .5894. That is completely nuts.

So the question is, why does women's hockey calculate its RPI so weirdly??

And in case you were wondering, the SOS calculation would change to:

1) Yale .6335
2) UMD .6251
3) SLU .5611
4) BSU .5509
5) Cornell .5487
6) OSU .5463
7) BC .5342
...
13) Minnesota .5055
...
28) Wisconsin .4868 (yikes)
 
Last edited:
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TTT, that is a very interesting analysis. One thing that jumps out to me is that all of the WCHA teams "benefit" in SOS by having played both UM and UW, while the Gophers and Badgers have not yet played each other, so their own SOS is skewed lower. It will be interesting to see how these numbers change after the UW-UM series, regardless of the outcome of that series.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TonyTheTiger20 said:
Just for kicks, here is what the RPI would look like if women's hockey used a standard .25/.50/.25 split instead of what is used now:

1) Wisconsin .7985
2) BC .7819
3) Minnesota .6854
4) Bemidji State .6335
5) Clarkson .6243
6) Quinnipiac .6180
7) Northeastern .5872
8) Minnesota-Duluth .5894 !!!!!!!
9) Harvard .5735
10) St. Lawrence .5842
11) Princeton .5720
12) North Dakota .5551

That is insane. Duluth's RPI goes from 18th in the country (.5048), and barely a TUC, all the way up to 8th!!! in the country at .5894. That is completely nuts.

So the question is, why does women's hockey calculate its RPI so weirdly??

And in case you were wondering, the SOS calculation would change to:

1) Yale .6335
2) UMD .6251
3) SLU .5611
4) BSU .5509
5) Cornell .5487
6) OSU .5463
7) BC .5342
...
13) Minnesota .5055
...
28) Wisconsin .4868 (yikes)
Click to expand...

Better, but I don't know if still smells right. Yale being number one? They played Minnesota and Q...that's about it. UMD on the otherhand has played UW, BC, UMN, Harvard, UND and BSU. Some of the rest (such as Cornell, BC), could be under debate...but they're in the realm I guess.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

5mn_Major said:
Better, but I don't know if still smells right. Yale being number one? They played Minnesota and Q...that's about it. UMD on the otherhand has played UW, BC, UMN, Harvard, UND and BSU. Some of the rest (such as Cornell, BC), could be under debate...but they're in the realm I guess.
Click to expand...
Oh I don't think just going to a standard .25/.50/.25 split would really 'fix' RPI or RPI's strength of schedule. It's still broken. It's more of an illustration of how changing an arbitrary set of coefficients can so drastically change things -- which suggests that it's a pretty crap way of doing things haha

Regardless, though, there is no doubt that Yale's strength of schedule belongs on the same level as UMD's right now.

First off:
They played Minnesota and Q...that's about it.
Click to expand...
That right there is 5 of Yale's 10 games haha

Yale's played one team outside of most people's top 15 -- one game against Merrimack. That's it.

Otherwise their entire schedule consists of BU (PWR 12th), Quinnipiac x3 (4th), Minnesota x2 (3rd), Princeton (10th), Harvard (6th), and Dartmouth (16th). That is a ridiculous schedule, and to say "that's about it" outside of Minnesota and Q shows some craaaaaaaaazy WCHA bias haha...

One bad team for Yale vs. four for UMD (Lindenwood x2 and Mankato x2). Despite the "top" of UMD's schedule being maybe a bit better than Yale's its bottom is definitely worse. Mathematically, that's where the difference comes from.

Anyway it's still wicked early for the Ivies -- those strength of schedule numbers aren't worth much for them right now. Neither UMD nor Yale's SOS is going to end up even close to 0.600, it's just not possible.
 
Last edited:
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

No one else but UMD has had 6 games against the top three teams plus an additional 2 against Harvard. Those 8 games "trump" the 5 of Yale to me.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

FiveHoleFrenzy said:
No one else but UMD has had 6 games against the top three teams plus an additional 2 against Harvard. Those 8 games "trump" the 5 of Yale to me.
Click to expand...
Duluth has played 16 games to Yale's 10. Both of those are half of its team's games. And anyway you can think Duluth's should be 1 and Yale's 2 or vice versa but I was just saying they both belong in the same stratosphere.
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

TonyTheTiger20 said:
Duluth has played 16 games to Yale's 10. Both of those are half of its team's games. And anyway you can think Duluth's should be 1 and Yale's 2 or vice versa but I was just saying they both belong in the same stratosphere.
Click to expand...

It's all good, I really just wanted to use the word trump in a sentence. ;)
 
Re: Fun With Numbers: 2016 Pairwise Predictor, What-Ifs, and Other Goodies

Has it been announced yet when the Women's Frozen Tickets go on sale?
 
You must log in or register to reply here.
Back
Top