PairWise Rankings 2010

[jimdahl]

Following up on a couple previous posts from this year...

When to start looking at PWR? (a data-based look at when PWR stabilizes)

First look at PWR contributor — non-conference records (in which I wonder aloud whether non-conference games matter as much more as we all think they do)

I just finished my first genuinely useful PWR analysis of the season:

PWR impact of Minnesota games vs. Cornell games

In that particular case (which is truly a single scenario that can't necessarily be extrapolated to answer the larger question), the games against Cornell prove no more important to North Dakota's current PWR ranking than the Minnesota games. An interesting, though not shocking, conclusion.

[sagard]

Nice stuff Jim. I don't pay as much attention (maybe it's because the Gophers have been so bad) as I used to, but your PWR analysis should be required reading for WCHA coaches or at least for staff. I think the comparison of the effects of the Cornell and the Gophers series leans more to the Gophers due to the fact that you aren't likely to lose many ECAC comparisons even with a split or worse vs. Cornell. Losing three points to the Gophers gives all your WCHA rivals a game and half boost on the CO category. I can't research it, but I'd guess a series with a top three CCHA team would have a larger impact on your PWR than the Gophers series.

Either way the key to RPI is clearly W/L record. From that follows PWR. It's really simple. Doesn't matter if you play a cupcakes or a tough schedule now, just win.

[jimdahl]

I think the comparison of the effects of the Cornell and the Gophers series leans more to the Gophers due to the fact that you aren't likely to lose many ECAC comparisons even with a split or worse vs. Cornell. Losing three points to the Gophers gives all your WCHA rivals a game and half boost on the CO category. I can't research it, but I'd guess a series with a top three CCHA team would have a larger impact on your PWR than the Gophers series.

Either way the key to RPI is clearly W/L record. From that follows PWR. It's really simple. Doesn't matter if you play a cupcakes or a tough schedule now, just win.

Yeah, I was just chatting with someone else about this via email. As I assured him, it's still a logically sound premise that all other things equal an NC game is more important than if the game were a conference game because the NC game contributes in a smaller COP pool (or paraphrasing his words, there are more opportunities to make up for a conference loss than an NC loss).

This outcome doesn't break that assumption, it just shows how easy it is for all other things not to be equal. In this case, I'm guessing that Minnesota manages to flip as many COP comparisons as Cornell. I'm still quite convinced that NC games are always more important than conference games because each one carries more individual weight in COPs, but I am starting to wonder if in reality that effect is so small that it doesn't matter much.

Nice stuff Jim. I don't pay as much attention (maybe it's because the Gophers have been so bad) as I used to, but your PWR analysis should be required reading for WCHA coaches or at least for staff.

This stuff does eventually sink in. I was pleased to see Brad manage to slip this into Rare visitor on tap (GF Herald):

There is another underlying reason why teams may be reluctant to travel to Ithaca for a nonconference game. The strength-of-schedule component has decreased in prominence in the Pairwise Rankings, which determine the 16-team NCAA tournament field.

...

Hakstol says that UND won’t make major philosophical changes in nonconference scheduling, though. His teams have made it a staple to play national powers.

“You can’t change the schedule just on a short-term basis,” he said. “We have contracts three, four years out. We are looking closely at it. There’s a combination there. We want to play good teams and make sure our fans have an opportunity to see good teams in our building. We want to make sure we’re challenged against some of the best in the country.

“At the same time, we want to make sure we’re giving ourselves a good opportunity to be in a good situation in the national tournament. We’re not going to make any drastic changes to that in the short term here.”

(for any relative newcomers who made it this far who don't know to what I'm referring, in this post I attempted to demonstrate that the NCAA's misguided '07 change to RPI simply served to weaken the importance of strength-of-schedule, which was definitely not the stated intent.)

[sagard]

I've pretty given up on looking at the other components of PWR, but I still check the RPI table for the Gophers and it looks pretty grim. To get to around 0.555 they need to win 9/12 and their first round series. That would likely get them in, anything else and they need to win the Final Five.

The WCHA dropped the ball allowing the RPI change to fly. I believe it has cost UW twice now and will continue to plague the fourth and fifth best WCHA teams.

[Goon]

I've pretty given up on looking at the other components of PWR, but I still check the RPI table for the Gophers and it looks pretty grim. To get to around 0.555 they need to win 9/12 and their first round series. That would likely get them in, anything else and they need to win the Final Five.

The WCHA dropped the ball allowing the RPI change to fly. I believe it has cost UW twice now and will continue to plague the fourth and fifth best WCHA teams.

You're right It allows teams like ECAC teams to get three bids, that takes from the other league like WCHA, HE and CCHA. I love watching teams like Clarkson bet a first round bid then get bounced their first game.

[The Sicatoka]

I've pretty given up on looking at the other components of PWR, but I still check the RPI table ...

I tend to look at KRACH because I believe that's a better model than the arbitrary weightings in an RPI.

I wish the NCAA would dump the RPI and use KRACH in the PWR. (That almost sounded dirty. )

[stickboy1956]

If we would have held on to win v Minn Friday we would be 7 in PWR right now.

With so many teams bunched together, 1 game/goal could be huge.

[AZSIOUX]

yeah these cornell games are big! go sioux

[sagard]

I tend to look at KRACH because I believe that's a better model than the arbitrary weightings in an RPI.

I wish the NCAA would dump the RPI and use KRACH in the PWR. (That almost sounded dirty. )

KRACH is cool when trying to get a feel for how closely you match up with an opponent, but obviously useless when trying to decide whether or not to buy regional tickets.

[jimdahl]

Finally managed to get the PWR forecaster up and running. Consider this a test run, as it really is the first time I've gotten it working this year, and there really may be something wrong.

If this is right, UND basically needs to sweep to avoid falling.

[brianvf]

That's brutal.

[xI Hammer Ix]

If that graph is right, which it is, things need to turn around tonight, no excuses, no more messing around, no more well "if/and/but this and that", must win situation.

Try not. Do, or do not. There is no try.

[Big A HG]

If that graph is right, which it is, things need to turn around tonight, no excuses, no more messing around, no more well "if/and/but this and that", must win situation.

Try not. Do, or do not. There is no try.

Well, so much for that, Hambone.

[jimdahl]

Finally managed to get the PWR forecaster up and running. Consider this a test run, as it really is the first time I've gotten it working this year, and there really may be something wrong.

If this is right, UND basically needs to sweep to avoid falling.

Well, as I warned, I didn't have full confidence in that chart.

Either something was quite wrong with it, or some very improbable outcomes really benefitted UND. I should be able to find out by digging through the predictions and comparing to the outcomes early this week.

UND at #10 in PWR

[sioux rube]

That loss to Alaska Anchorage last nite hurt CC pretty good.

[jimdahl]

Well, as I warned, I didn't have full confidence in that chart.

Either something was quite wrong with it, or some very improbable outcomes really benefitted UND. I should be able to find out by digging through the predictions and comparing to the outcomes early this week.

UND at #10 in PWR

I wish I could say what went wrong with that first run, but I don't really know. Here's what the forecast should have been:

[The Whistler]

I wish I could say what went wrong with that first run, but I don't really know. Here's what the forecast should have been:

It's a lot easier waiting until Monday to make your prediction.

I think the vikes are going to lose in overtime.

[Oxbow6]

It's a lot easier waiting until Monday to make your prediction.

I think the vikes are going to lose in overtime.

In my best Paul Allen voice..."Noooooooooooooooooooooooo!"

[siouxweet]

In my best Paul Allen voice..."Noooooooooooooooooooooooo!"

I would love to hear his call on the ot kick last night.

[Oxbow6]

I would love to hear his call on the ot kick last night.

As a Packer fan, I was laughing so hard at his call after the Favre INT at the end of regulation...it's awesome!

[siouxweet]

As a Packer fan, I was laughing so hard at his call after the Favre INT at the end of regulation...it's awesome!

I heard his call on the third AP fumble when he started to call for him to be taken out of the game, that was priceless

edit: today me as a bear fan and you as a packer fan have something to agree on

[Oxbow6]

I heard his call on the third AP fumble when he started to call for him to be taken out of the game, that was priceless

edit: today me as a bear fan and you as a packer fan have something to agree on

But only for today! And Culter is still a bum!

[siouxweet]

But only for today! And Culter is still a bum!

only for today

[jimdahl]

It's a lot easier waiting until Monday to make your prediction.

If only I had waited until Monday, then I would have gotten it right

Anyway, now that it's correctly forecasting backward, I'm actually quite a bit more confident in its results looking forward. Here's a chart:

And here are the key games:

Merrimack over New Hampshire

Ohio State over Ferris State (sweep is a much bigger bump than a split)

St Cloud St over CC (sweep is a much bigger bump than a split)

Michigan over Mich St.

Mass. losing to BU and Providence

For those most part those look like RPI flips UND PWR Details

[jimdahl]

Still feeling good about how well the simulator seems to be working, a prediction of Minnesota for our Gopher friends:

A sweep guarantees staying a TUC. A split makes staying a TUC very likely. Don't get swept.