Bracketology 2016-17 Style

jkahn

BearLover

Beeeej

Trotsky
Latest odds from PlayoffStatus.com:

ECAC Frozen Four: 75%
ECAC Title Game: 38%
ECAC Champions: 18% (1 in 6)

Make the NCAA tourney: 91%
Round 2: 43%
Frozen Four: 19% (1 in 5)
Title Game: 9% (1 in 11)
NCAA Champions: 4% (1 in 25)

I'll take those odds. After 36 seasons as a fan they don't seem daunting.

But Greg, those odds don't take into account the bad bounces that frequently occur in hockey games, or the possibility that a giant space squid will crash down through the roof of Lynah Rink and land on Mitch Gillam. Why do you hate America?

You joke, but those odds look far more realistic than the probability matrix ones. 75% chance of making Lake Placed is considerably more reasonable than 95% or whatever the other model gave us.

This 91% for us making the tournament or the 98% shown on CHN
[www.collegehockeynews.com]
seem way high to me (and by "to me" I mean my quick analysis using KRACH probabilities).
Per Krach, there's a 27.5% chance we don't get by Clarkson this weekend. If we lose that series 2-1, our winning percentage drops to .6719. To try to see that effect, using CHN's pairwise calculator, I simply flipped one of our wins (I used the UNH game) to a loss, which drops are current percentage to .6724. If that were the case, we'd be at #14 in pairwise, with a fair amount of risk. And there's an 11.9% chance of getting swept, which is obviously worse. I realize that this is overly simplistic and ignores all other results happening around us.
I hope these models are accurate, but more importantly, let's take care of business this weekend.

One thing to keep in mind is that there will be a bunch of losses for other teams in contention. Many of them will play each other, sometimes in best of threes. Dropping down to 14th, all else being equal, still gives us a pretty decent shot when you factor in how the teams around us will do.

Agreed about taking care of business.

jkahn

BearLover

Beeeej

Trotsky
Latest odds from PlayoffStatus.com:

ECAC Frozen Four: 75%
ECAC Title Game: 38%
ECAC Champions: 18% (1 in 6)

Make the NCAA tourney: 91%
Round 2: 43%
Frozen Four: 19% (1 in 5)
Title Game: 9% (1 in 11)
NCAA Champions: 4% (1 in 25)

I'll take those odds. After 36 seasons as a fan they don't seem daunting.

But Greg, those odds don't take into account the bad bounces that frequently occur in hockey games, or the possibility that a giant space squid will crash down through the roof of Lynah Rink and land on Mitch Gillam. Why do you hate America?

You joke, but those odds look far more realistic than the probability matrix ones. 75% chance of making Lake Placed is considerably more reasonable than 95% or whatever the other model gave us.

This 91% for us making the tournament or the 98% shown on CHN
[www.collegehockeynews.com]
seem way high to me (and by "to me" I mean my quick analysis using KRACH probabilities).
Per Krach, there's a 27.5% chance we don't get by Clarkson this weekend. If we lose that series 2-1, our winning percentage drops to .6719. To try to see that effect, using CHN's pairwise calculator, I simply flipped one of our wins (I used the UNH game) to a loss, which drops are current percentage to .6724. If that were the case, we'd be at #14 in pairwise, with a fair amount of risk. And there's an 11.9% chance of getting swept, which is obviously worse. I realize that this is overly simplistic and ignores all other results happening around us.
I hope these models are accurate, but more importantly, let's take care of business this weekend.

PlayoffStatus also notes that an 0-2 record this weekend would drop CU's chances to 61%

i was at the canisus games in oct where they lost 1-0 and 2-1 and all 3 goals came because of bad bounces, simple passes and pucks that hopped over sticks at the blue line turned into breakaways 3 times in games they outshot them 85-40.. cant blame it all on puck luck for sure. but they were snake but with that and injuries this year. he thinks he passed on the cornell lack of scoring MOJO to RIT..

upprdeck
i was at the canisus games in oct where they lost 1-0 and 2-1 and all 3 goals came because of bad bounces, simple passes and pucks that hopped over sticks at the blue line turned into breakaways 3 times in games they outshot them 85-40.. cant blame it all on puck luck for sure. but they were snake but with that and injuries this year. he thinks he passed on the cornell lack of scoring MOJO to RIT..

great typo! He said 'but' heh..heh...heh

This week’s brackets

East Regional (Providence):

13 Wisconsin vs. 4 Western Michigan
11 Providence vs. 8 Union

Northeast Regional (Manchester):

14 Notre Dame vs. 3 Harvard
10 Cornell vs. 6 UMass Lowell

Midwest Regional (Cincinnati):

15 Canisius vs. 2 Minnesota Duluth
9 Penn State vs. 7 Boston University

West Regional (Fargo):

16 Bemidji State vs. 1 Denver
12 North Dakota vs. 5 Minnesota

Conference breakdowns

NCHC — 4
Hockey East — 4
ECAC Hockey — 3
Big Ten — 3
WCHA — 1
Atlantic Hockey — 1

Movement

In: North Dakota
Out: Ohio State

Read more: [www.uscho.com]

I like this one.

Jim Hyla
I like this one.

Me too. Bring it on.

any bracket that has us in it is good to me.

Not to mention the possible quarterfinal matchup with Harvard.

Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

I sincerely hope you've got "the right stuff".

Jim Hyla

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

I sincerely hope you've got "the right stuff".

I'm happy to answer the actual question, since I'm sure he's not expecting a serious one:

Yes, the statistical probabilities factor in every variable you mention, because the prior game results on which the statistical probabilities are based resulted from those variables, among many others.

But again, we play the actual games, because there is no telling what might happen on any given night. If you get all your chips in on the turn in a huge poker hand, and the other player has only one out in the deck remaining to beat you, you have a roughly 98% chance of winning the hand. If you lose the hand because that one card comes out on the river, it doesn't mean the statistical probabilities were wrong.

(Disclaimer: An example based purely on luck is not intended to imply that the outcome of a hockey game is based purely on luck, but luck is certainly one factor.)

Beeeej

Jim Hyla

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

I sincerely hope you've got "the right stuff".

I'm happy to answer the actual question, since I'm sure he's not expecting a serious one:

Yes, the statistical probabilities factor in every variable you mention, because the prior game results on which the statistical probabilities are based resulted from those variables, among many others.

But again, we play the actual games, because there is no telling what might happen on any given night. If you get all your chips in on the turn in a huge poker hand, and the other player has only one out in the deck remaining to beat you, you have a roughly 98% chance of winning the hand. If you lose the hand because that one card comes out on the river, it doesn't mean the statistical probabilities were wrong.

(Disclaimer: An example based purely on luck is not intended to imply that the outcome of a hockey game is based purely on luck, but luck is certainly one factor.)

1+

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

How about a big hand for Oscar Wilde, ladies and gentlemen?

Beeeej

Jim Hyla

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

I sincerely hope you've got "the right stuff".

I'm happy to answer the actual question, since I'm sure he's not expecting a serious one:

Yes, the statistical probabilities factor in every variable you mention, because the prior game results on which the statistical probabilities are based resulted from those variables, among many others.

But again, we play the actual games, because there is no telling what might happen on any given night. If you get all your chips in on the turn in a huge poker hand, and the other player has only one out in the deck remaining to beat you, you have a roughly 98% chance of winning the hand. If you lose the hand because that one card comes out on the river, it doesn't mean the statistical probabilities were wrong.

(Disclaimer: An example based purely on luck is not intended to imply that the outcome of a hockey game is based purely on luck, but luck is certainly one factor.)

Okay everyone let me handle this one...

If you can't or refuse to quantify key variables you have no business making statistical predictions, particularly predictions expressed in fractions of a percent. Are all you statistical savants hotel administration majors?

ugarte

Beeeej

Jim Hyla

ugarte

Hooking
Do statistical probabilities factor in variables such as player conditioning, team discipline on and off the ice, desire, dedication, coaching skills, game plans and adaptability - interesting stuff like that? I hope so.

OK everyone let me handle this one.

Hooking, shut up already.

I sincerely hope you've got "the right stuff".

I'm happy to answer the actual question, since I'm sure he's not expecting a serious one:

Yes, the statistical probabilities factor in every variable you mention, because the prior game results on which the statistical probabilities are based resulted from those variables, among many others.

But again, we play the actual games, because there is no telling what might happen on any given night. If you get all your chips in on the turn in a huge poker hand, and the other player has only one out in the deck remaining to beat you, you have a roughly 98% chance of winning the hand. If you lose the hand because that one card comes out on the river, it doesn't mean the statistical probabilities were wrong.

(Disclaimer: An example based purely on luck is not intended to imply that the outcome of a hockey game is based purely on luck, but luck is certainly one factor.)

Okay everyone let me handle this one...

I'm willing, but it doesn't look like it worked the first time, so try harder, will you please. Pretty please.

Hooking
If you can't or refuse to quantify key variables you have no business making statistical predictions, particularly predictions expressed in fractions of a percent. Are all you statistical savants hotel administration majors?

OK. Let me try this time.

Look, you don't have to include every individual contributing factor separately in order to have a viable model. There are such things as interaction effects, synergies, and emergent properties so that the whole is more than the sum of its parts. You just have to have a model that incorporates effects that result from all the underlying factors and their interactions.

On top of this, many of the contributing factors are themselves correlated or, more precisely, different aspects of the same thing. So including them individually contributes nothing, and in some ways may even make the model weaker.

Finally, let's not forget that the world does not consist of "variables." For the most part, these are just measurements of the underlying reality. The reality itself is interconnected in a single whole. The measurements treated as "variables" may help us understand the reality, unless we make the mistake of treating this pragmatic, epistemic practice as if it had ontological status outside and independent of that practice.

Since someone mentioned hotel administration majors, I'll recapitulate these points through a more pedestrian example. No doubt, the quality of a great meal depends on the genetic makeup of the ingredients, how the farmers grew them, how fresh and local the ingredients are, the skills of the cooks, the quality of the kitchen appliances and utensils, the imagination of the menu planner, the knowledge and skill of the chef who invents various unique dishes, and so on. But for the most part, one does not have to know all these details to be confident of having a great meal. Usually the general reputation of the restaurant is sufficient for, say, 75% confidence. Past experience eating there might add 20%. Recommendation from a trusted friend, another 15%. And a positive review by a reliable food critic, 10 - 15% more. As you can see from the probabilities, even some of these may be redundant.

I can't believe you are still responding to him in good faith.

ugarte
I can't believe you are still responding to him in good faith.

+1

No offense to hotel school majors. The captain of the first Cornell Men's Hockey Team to play in Lynah, Roger ("the beast" Eastman, was a hotel major. Sorry, I don't have the statistics on Roger.

Hooking
No offense to hotel school majors. The captain of the first Cornell Men's Hockey Team to play in Lynah, Roger ("the beast" Eastman, was a hotel major. Sorry, I don't have the statistics on Roger.

Looks like nobody does:

[www.hockeydb.com]

Or maybe they just decided not to publish his stats because they wouldn't accurately reflect what he ate for lunch on game days.

Hooking
The captain of the first Cornell Men's Hockey Team to play in Lynah, Roger ("the beast" Eastman, was a hotel major. Sorry, I don't have the statistics on Roger.

[www.tbrw.info]

Thanks. Those statistics explain why we lost to the Susquehanna Club Team and we didn't make it to the NCAA finals. Wait! the statistical fact that the '58 Cornell hockey team was not in Division I means there was a 100% chance Cornell could not win the NCAA D-1 championship - just as certain as the statistical fact that if today's Cornell team doesn't keep winning they won't be 2017 D-1 NCAA men's hockey champions, all other statistics notwithstanding.

Hooking
just as certain as the statistical fact that if today's Cornell team doesn't keep winning they won't be 2017 D-1 NCAA men's hockey champions, all other statistics notwithstanding.

Actually Cornell can definitely lose one, and perhaps even two more games and still wind up as 2017 D-1 NCAA Men's Hockey Champions.

What are the statistical odds?

Hooking
Wait! the statistical fact that the '58 Cornell hockey team was not in Division I means there was a 100% chance Cornell could not win the NCAA D-1 championship

There was no D-1 in 1958 and AFAIK Cornell was eligible for the NC$$ tournament. Not sure why they were passed over with their gaudy 3-7-1 record.

At some point you need to let stoopid people be stoopid and get over it.

Hooking, let us discuss statistics in our own, flawed way. If you don't like it, choose another thread. Or forum.

Everyone else, don't feed the trolls.

We are a day away from hockey... can't wait!

Sound advice, Ken. I will read comments about Cornell Ice Hockey and try to ignore statistical speculation, no matter how annoying it might be.

just looping thru the college hockey news site and using the rpi tool it was pretty easy to find ways that we would still be in after losing both games this weekend and thats without factoring in that more teams will lose again after that. just root for chalk in most of the other leagues and 3 game series in the ones that have teams both behind us.

upprdeck
just looping thru the college hockey news site and using the rpi tool it was pretty easy to find ways that we would still be in after losing both games this weekend and thats without factoring in that more teams will lose again after that. just root for chalk in most of the other leagues and 3 game series in the ones that have teams both behind us.

I'm thinking our ideal is all the ECAC QF (in particular, US) are 2-0 home. That sets up a Placid field that could be stronger than some NC$$ regionals.

thats why if we got 2-0 its very hard for us to lose next week 1 game and fall out, several teams around us have to lose twice. winning 2-1 and then losing becomes much dicier.

So - this conversation is actually quite relevant given that Cornell's odds have dropped to 65% in the new Matrix after today's game .... I have many thoughts on this topic, but no real answers. I appreciate constructive discussion.

[www.collegehockeynews.com]

* Going into the weekend, Cornell, Penn State and Providence all had roughly 95%+ odds of making the NCAAs. Penn State and Providence both lost Friday, but barely budged. Cornell, on the other hand, dropped quite a bit. To approximately 65% ... Why is this? I wish I definitively knew the answer to this. My math skills are OK, but not sure I have all the tools to properly answer that question. Intuitively, I wonder if it's because one particular flaw of the Matrix model that I point out on the site ... that it doesn't recalculate the KRACH after each game in each iteration. That would be ridiculously problematic and take forever to run. But a Cornell loss/Clarkson win may have shifted things enough to drop the odds of winning Saturday by a good amount. I don't know, just a theory.

* I don't think the KRACH model is flawless by any means, but mostly I'd attribute whatever flaws there are to small sample size and a bit of circular logic that makes it work. Of course it can't take into account all factors - what can? But that's only because of sample size. The intangible factors described above are baked into the existing results. The mathematicians among us can explain this a lot better than I can - but it's about confidence intervals and what not. No one says the data is flawless. No one suggests the games don't have to be played on the ice. Of course they do. But a system doesn't have to be flawless in order to give a sound assessment of odds of finishing in a certain spot.

* KRACH is actually not meant to be predictive. It's meant to be descriptive of what's already taken place. It's much closer to ideal in that direction. I think that's probably obvious to most, but worth reminding. I think KRACH is generally the best model I've seen. That said, I have no problem if someone has an honest opinion about not buying certain odds being what they are. It just depends on what the argument is. If it's dismissive of all math, then the argument has no merit. But there is room to argue on the edges of how these things can be better. I'm all ears.

adamw
So - this conversation is actually quite relevant given that Cornell's odds have dropped to 65% in the new Matrix after today's game .... I have many thoughts on this topic, but no real answers. I appreciate constructive discussion.

[www.collegehockeynews.com]

* Going into the weekend, Cornell, Penn State and Providence all had roughly 95%+ odds of making the NCAAs. Penn State and Providence both lost Friday, but barely budged. Cornell, on the other hand, dropped quite a bit. To approximately 65% ... Why is this? I wish I definitively knew the answer to this. My math skills are OK, but not sure I have all the tools to properly answer that question. Intuitively, I wonder if it's because one particular flaw of the Matrix model that I point out on the site ... that it doesn't recalculate the KRACH after each game in each iteration. That would be ridiculously problematic and take forever to run. But a Cornell loss/Clarkson win may have shifted things enough to drop the odds of winning Saturday by a good amount. I don't know, just a theory.

Glad you brought this up. I've had some thoughts for how well the matrix is doing.
Yesterday, the matrix gave us a 98% chance to make the NCAAs. Our probability of being swept was around 13% (probably a little higher since our KRACH would go down for our second game). Assuming we only don't make it if we get swept, that means 11% of the 13% we get swept, we still make it in. That's over an 80% chance of getting in while getting swept.

Today, we have a 61% chance of winning tonight's game and a 65% chance of making the NCAAs. Making the same assumption as above, we would only make the tournament 4% out of the 39% when we got swept (about 11%). Obviously, this is too high. There are probably numerous scenarios we don't make it if we lose in three.

Back to your question, why the big drop? Maybe we're going down that 2% path based on what's going on with other teams. It's tough to judge a predictive model on just one year of data. I look forward to more years of seeing how well it does.

KGR11

adamw
So - this conversation is actually quite relevant given that Cornell's odds have dropped to 65% in the new Matrix after today's game .... I have many thoughts on this topic, but no real answers. I appreciate constructive discussion.

[www.collegehockeynews.com]

* Going into the weekend, Cornell, Penn State and Providence all had roughly 95%+ odds of making the NCAAs. Penn State and Providence both lost Friday, but barely budged. Cornell, on the other hand, dropped quite a bit. To approximately 65% ... Why is this? I wish I definitively knew the answer to this. My math skills are OK, but not sure I have all the tools to properly answer that question. Intuitively, I wonder if it's because one particular flaw of the Matrix model that I point out on the site ... that it doesn't recalculate the KRACH after each game in each iteration. That would be ridiculously problematic and take forever to run. But a Cornell loss/Clarkson win may have shifted things enough to drop the odds of winning Saturday by a good amount. I don't know, just a theory.

Glad you brought this up. I've had some thoughts for how well the matrix is doing.
Yesterday, the matrix gave us a 98% chance to make the NCAAs. Our probability of being swept was around 13% (probably a little higher since our KRACH would go down for our second game). Assuming we only don't make it if we get swept, that means 11% of the 13% we get swept, we still make it in. That's over an 80% chance of getting in while getting swept.

Today, we have a 61% chance of winning tonight's game and a 65% chance of making the NCAAs. Making the same assumption as above, we would only make the tournament 4% out of the 39% when we got swept (about 11%). Obviously, this is too high. There are probably numerous scenarios we don't make it if we lose in three.

Back to your question, why the big drop? Maybe we're going down that 2% path based on what's going on with other teams. It's tough to judge a predictive model on just one year of data. I look forward to more years of seeing how well it does.

I've felt all along that the model was giving us way too high a percentage chance, and previously posted that our with a 27.5% KRACH chance of losing in the quarters and knowing our RPI would drop way more than others with losses due to our high winning percentage, the models just didn't seem accurate. Without really understanding the logic used in the models, I can't comment further, other than I don't think that readjusting KRACH would have made that big a difference.

pwr is a strange beast. Prov loses and gains pwr, penn state loses to a team worse than clarkson and goes no where.

vt/bc hard to tell which is better for us but i have a feeling a 2-1 series helps us the most.

wisc over osu might help us tonight.

we win 2 and we are in good shape if chalk wins the hockey east and b10.

It's our win percentage. A single loss hurts us way more than other teams that are around us in RPI with lower win percentages.

Blame brown, RPI, and Colgate.

I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

BearLover
I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

A lot of that is that we've overachieved this year, in terms of what our win% "should" be compared to our goal differential. We've won a lot of close games.

At the same time, Clarkson is pretty good. The high chances of making the tournament have more to do with the fact that we have a pretty good shot if we don't win in this round. That being said, it looks like yesterday was a disaster for us in terms of other results, not to mention the actual disaster last night.

So, let's win a few.

BearLover
I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

Yes and no. I'm not sure I'd use "regression to the mean" as the right way to put it, because that assumes you know what the mean is, which you really can't do from past results. But I know what you're driving at. Intuitively you think Cornell-Clarkson were closer. But they key is, how do you model that? Not just go by "feel." I think there may be a way to use goal differential and things like PDO and Team Corsi, to come up with some sort of counter-weight on straight KRACH. But I couldn't tell you exactly what that would be. Goal diffs have always been a dicey thing in hockey. Corsi may be better, but has its flaws. There might be some balance there.

I'd also like to be able to definitively answer the question why Cornell was affected more than Penn State, for example. With an actual demonstration.

adamw

BearLover
I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

Yes and no. I'm not sure I'd use "regression to the mean" as the right way to put it, because that assumes you know what the mean is, which you really can't do from past results. But I know what you're driving at. Intuitively you think Cornell-Clarkson were closer. But they key is, how do you model that? Not just go by "feel." I think there may be a way to use goal differential and things like PDO and Team Corsi, to come up with some sort of counter-weight on straight KRACH. But I couldn't tell you exactly what that would be. Goal diffs have always been a dicey thing in hockey. Corsi may be better, but has its flaws. There might be some balance there.

I'd also like to be able to definitively answer the question why Cornell was affected more than Penn State, for example. With an actual demonstration.

There's a flaw in the model somewhere. If a loss last night brings Cornell to a 65% NCAA chance, and there was a 35% KRACH chance of the happening, and if we assume that the other results were pretty much at an average expectation, then:
if the was a 35% chance of 65% after the game and a 65% chance of 98% or higher (say even 100%), then the chance before last night should be no greater than .35 x .65 + .65 x 1.00 or about 88%. And even the 65% after last night feels way overstated to me (again based on KRACH). Using the KRACH prior to last night, there's now a 35% chance we'll be swept and the odds are now against us winning the series. And if we lose in three, the added 1-1 to our record will drop our RPI further.

adamw

BearLover
I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

Yes and no. I'm not sure I'd use "regression to the mean" as the right way to put it, because that assumes you know what the mean is, which you really can't do from past results. But I know what you're driving at. Intuitively you think Cornell-Clarkson were closer. But they key is, how do you model that? Not just go by "feel." I think there may be a way to use goal differential and things like PDO and Team Corsi, to come up with some sort of counter-weight on straight KRACH. But I couldn't tell you exactly what that would be. Goal diffs have always been a dicey thing in hockey. Corsi may be better, but has its flaws. There might be some balance there.

I'd also like to be able to definitively answer the question why Cornell was affected more than Penn State, for example. With an actual demonstration.

To the extent Cornell (or any team) has over-achieved against expectations you're going to see these weird scenarios in any model. The issue is one of modeling uncertainty. You know the input - raw game outcomes - is information poor given the length of the college hockey season. Your model, therefore, has fairly high uncertainty baked into it. If you just do iterative monte carlo you're going to get a particular distribution of results but that doesn't mean you can take the probability output and say "Cornell is 90% likely" - you need to understand where you might be affected by hidden error and apply other measures to balance the outcome predictions the model uses. Basically, straight KRACH is not sufficiently information rich to avoid this kind of overstated certainty.

This is a big reason why the 538 election predictor had Trump at 30% to win and 10% to win while losing the popular vote when the Princeton Election model had a 98%+ chance for a Clinton victory. 538 assumed, based on past polling data, that polling errors tend to be correlated across clusters of states, and applying that correction shifted the 2016 prediction drastically in ways that it did not in 2012, when correlated error wasn't likely to matter.

To go back to hockey, KRACH helps account for the uncertainty caused by insular schedules which either unfairly boost records of teams in weak conferences or unfairly punish teams in strong conferences. I mean, that's pretty much what it does. However, it doesn't account for the uneven distribution of random outcomes you're likely to see over a small sample of college teams playing a short schedule. I don't know that reversion to the mean is the explanation in this case, but it's an explanation, and we can get a better model of what that "mean" should be than whatever current record would indicate. Like you, I'm not sold on goal differential in general but looking at fraction of 1-goal outcomes might help. I do think PDO and corsi are worth a look.

Something to consider:

Cornell's PDO this season is ~102. Corsi around 50%
Penn State: PDO is ~98. Corsi around 60%
Providence: PDO is ~100. Corsi around 56%

Let's assume, and I know not everybody around here agrees, that these shot attempt rates are predictive of game success over a large sample but have high variance and low predictive certainty for individual games or even entire NCAA seasons.

Now consider schedule strength rankings and schedule outcomes - these three teams are pretty even on that front. But assuming those season averages for Corsi/PDO are representative, and not being dragged way up or down by a handful of outliers, that suggests Cornell's tournament prediction % should have been substantially lower than the predictions for Penn State or Providence from the beginning. Cornell's record was a lot better than one might expect relative to schedule strength and on-ice performance as reflected in these metrics. The pure KRACH-based model clobbers Cornell for losing a "should win" game because the next iteration both accounts for the damage to Cornell's record caused by the loss *and* the sharper decline in predicted win rate as a result of the fall in past win %. In this case, at least, if you use just a corsi weighting adjustment to the KRACH prediction Cornell's predicted chances probably would've dropped drastically from the outset, because there would've been a downward adjustment to their raw KRACH-based win probabilities. The model would almost certainly be less volatile with respect to this specific comparison set.

Now, that doesn't mean the model would be better, and how to adjust it correctly is the hard part - there are whole bodies of literature about this. One way is to do it empirically - you've got a lot of college hockey playoff history to use, and you can simulate all manner of different models and test their predictions against the observed distributions. Even just 10 years * N teams where N is number of conference tournament QF participants gives you something to model against. It's worth investigating in the offseason - if you've already got the shot data it's not going to be all that hard.

jkahn

adamw

BearLover
I don't know enough about the model/KRACH/etc. to say anything especially productive, but I will say that, as a casual observer, the 98% number, the 91% number from a different model, and the 85% number from the Matrix before the RPI game all failed the eye test. If I had to guess why, it's because KRACH, as adamw said, is meant to be descriptive of what has occurred rather than predictive. Thus, it doesn't account for regressions to the mean, etc. Cornell is not going to beat RPI nine times out of ten, even if their past records would equate to such a mismatch. And beating Clarkson is far closer to a coin flip than a sure thing, even though the models gave Cornell a very high chance of winning that too.

Yes and no. I'm not sure I'd use "regression to the mean" as the right way to put it, because that assumes you know what the mean is, which you really can't do from past results. But I know what you're driving at. Intuitively you think Cornell-Clarkson were closer. But they key is, how do you model that? Not just go by "feel." I think there may be a way to use goal differential and things like PDO and Team Corsi, to come up with some sort of counter-weight on straight KRACH. But I couldn't tell you exactly what that would be. Goal diffs have always been a dicey thing in hockey. Corsi may be better, but has its flaws. There might be some balance there.

I'd also like to be able to definitively answer the question why Cornell was affected more than Penn State, for example. With an actual demonstration.

There's a flaw in the model somewhere. If a loss last night brings Cornell to a 65% NCAA chance, and there was a 35% KRACH chance of the happening, and if we assume that the other results were pretty much at an average expectation, then:
if the was a 35% chance of 65% after the game and a 65% chance of 98% or higher (say even 100%), then the chance before last night should be no greater than .35 x .65 + .65 x 1.00 or about 88%. And even the 65% after last night feels way overstated to me (again based on KRACH). Using the KRACH prior to last night, there's now a 35% chance we'll be swept and the odds are now against us winning the series. And if we lose in three, the added 1-1 to our record will drop our RPI further.

Keep in mind that other games factor in, too. It may be that yesterday's other results were unlikely and/or bad for us, skewing the numbers.

Tom Lento
Now, that doesn't mean the model would be better, and how to adjust it correctly is the hard part - there are whole bodies of literature about this. One way is to do it empirically - you've got a lot of college hockey playoff history to use, and you can simulate all manner of different models and test their predictions against the observed distributions. Even just 10 years * N teams where N is number of conference tournament QF participants gives you something to model against. It's worth investigating in the offseason - if you've already got the shot data it's not going to be all that hard.

Tom, all of this perfectly explains why KRACH is not a good predictive model (or, good enough) and I greatly appreciate the discussion on ways the simulation can be improved.

But this still doesn't necessarily answer why Cornell dropped as much as they while other teams did not. Or maybe you did explain it, and I'm not getting it.

In other words - there's no way for KRACH to know, before hand, that Cornell has theoretically over-achieved. So while that might explain the flaw in its predictive value - it doesn't explain the math of why they dropped when others didn't.

Am I making sense?

By the way - we don't have Corsi-esque shot data before 2 seasons ago. We have general shots per game going back to 2002, but that's not as good.

adamw

Tom Lento
Now, that doesn't mean the model would be better, and how to adjust it correctly is the hard part - there are whole bodies of literature about this. One way is to do it empirically - you've got a lot of college hockey playoff history to use, and you can simulate all manner of different models and test their predictions against the observed distributions. Even just 10 years * N teams where N is number of conference tournament QF participants gives you something to model against. It's worth investigating in the offseason - if you've already got the shot data it's not going to be all that hard.

Tom, all of this perfectly explains why KRACH is not a good predictive model (or, good enough) and I greatly appreciate the discussion on ways the simulation can be improved.

But this still doesn't necessarily answer why Cornell dropped as much as they while other teams did not. Or maybe you did explain it, and I'm not getting it.

In other words - there's no way for KRACH to know, before hand, that Cornell has theoretically over-achieved. So while that might explain the flaw in its predictive value - it doesn't explain the math of why they dropped when others didn't.

Am I making sense?

By the way - we don't have Corsi-esque shot data before 2 seasons ago. We have general shots per game going back to 2002, but that's not as good.

I'm almost certain it's our higher win%, which means one loss hurts our RPI more than other teams. I can't dig up the RPI numbers from before yesterday, but CHN lets me set individual games to ties. We're at .5455. Had we tied (I know, impossible), we'd be at .5499. Had we won, .5543.

Doing the same with Penn State and Providence, Penn State is at .5507. Had they tied, .5546, had they won, .5584. Smaller boosts than we would have had, but significant.

Providence is at .5516, had they tied, .5554, had they won, .5589. So, same as Penn State, roughly.

The results are less striking than I expected. Weirdly, we shuffled up to 12th in some of these scenarios.

In conclusion, I don't have much of a conclusion.

How about the simplest explanation- we could just have had a number of PWR comparisons that were razor close? THat matrix uses krach to simulate game results but ultimately fills out the PWR- so the issue is more likely in the PWR than the karch numbers.

Looking at the probability matrix, I think you need to look at the probabilities of landing in any given PWR final position. I'm curious what the distribution was for us across positions before last night, because as of 920 pm, we have crazy high probabilities of landing in 15 or 16, while Penn St. and Providence skew their individual probabilities much more towards the higher PWR finishes

The answer to the riddle will be found by seeing which individual PWR comparisons are likely to get flipped against us as opposed to the other two teams having the same thing happen.

adamw

Tom Lento
Now, that doesn't mean the model would be better, and how to adjust it correctly is the hard part - there are whole bodies of literature about this. One way is to do it empirically - you've got a lot of college hockey playoff history to use, and you can simulate all manner of different models and test their predictions against the observed distributions. Even just 10 years * N teams where N is number of conference tournament QF participants gives you something to model against. It's worth investigating in the offseason - if you've already got the shot data it's not going to be all that hard.

Tom, all of this perfectly explains why KRACH is not a good predictive model (or, good enough) and I greatly appreciate the discussion on ways the simulation can be improved.

But this still doesn't necessarily answer why Cornell dropped as much as they while other teams did not. Or maybe you did explain it, and I'm not getting it.

In other words - there's no way for KRACH to know, before hand, that Cornell has theoretically over-achieved. So while that might explain the flaw in its predictive value - it doesn't explain the math of why they dropped when others didn't.

Am I making sense?

By the way - we don't have Corsi-esque shot data before 2 seasons ago. We have general shots per game going back to 2002, but that's not as good.

Drat, I thought the data went back at least 4-5 seasons before going into shot per game territory.

This isn't at all about KRACH "knowing" about the fact that Cornell over-achieved. It's just what you might see when a model suddenly hits a sharp correction at some extreme point in the distribution. The fact that the loss pulled Cornell's record in the direction of what corsi would suggest was as much coincidence as anything. The real question is whether or not such a volatile result is reasonable.

The cause of this volatility for Cornell but not Penn State is much harder to tease apart, and frankly I didn't really answer your question directly because I can't. One hypothesis is others here are right and the reason has to do with relative winning percentages and either how that interacts with RPI or some artifact of the KRACH-based predictive model that tends to over-emphasize high winning percentages. That's really the only difference I notice between those teams - they all have similar outcomes against reasonably similar competition (Providence less so than Penn State), but Cornell has a much higher win %. I'm not an expert here so I'm really guessing, but this makes intuitive sense to me when I think about how KRACH handles perfect teams - they're expected to be perfect against everybody, and as soon as they drop a point to somebody their rating falls sharply into line with a more reasonable projection.

You could test this win % hypothesis - run the model on Thursday's inputs. Then do it again but update Cornell's record to include a 2-2-1 season series against an imaginary KRACH-neutral opponent. See how it changes the model predictions. See how much Cornell's volatility changes after Friday's results are added to the model. If the volatility is due to the model having a self-reinforcing high win % expectation, you'd expect 1) Cornell's odds on the pre-Friday model to decrease and 2) Cornell's shift in odds on the post-Friday model to be less volatile.

Of course it could also just be related to Cornell's relative position in the comparison rankings.

all it has shown is that at 2-0 we were sitting pretty good. anything less was killer.

the question will be what happen when the b10 teams beat up on each other next weekend

do we fall behind idle Vt if we lose

does 1-1 get us past ND if they lose are we rooting for stsloud to come back against nDakota

Miami blows it late to a top team one last time.

St. Cloud and ND going to OT. I think we want St. Cloud, after all. At the very least, we want them to win tonight and take ND to 3 games.

Two things strangely missing from this rather technical discussion are standard errors and confidence intervals. I'm neither familiar with, nor particularly interested in, the inner workings of the models. But it seems to me that, other things being equal, the standard errors of a prediction regarding a team's prospects will decrease as the team plays more games, but the impact will decrease as the number of games increases. Since Cornell has played fewer games than the factory schools, one would therefore expect an additional game to have a bigger impact on Cornell's standard errors than the competition's. Also, since confidence intervals are wider with smaller N, an additional datum will have a larger impact, percentage wise.

But since such considerations have been absent from the discussion thus far, I have the impression that the predictions are essentially point estimates rather than interval estimates (with some plausibly associated probability distribution) -- the famous "margin of error" routinely misrepresented and misunderstood by the media when reporting survey data.

Comments from someone in the know?

Swampy
But since such considerations have been absent from the discussion thus far, I have the impression that the predictions are essentially point estimates rather than interval estimates (with some plausibly associated probability distribution) -- the famous "margin of error" routinely misrepresented and misunderstood by the media when reporting survey data.

Comments from someone in the know?

This is generally correct, yes. The prediction is made based on a set of simulation results, which produces a frequency distribution of outcomes reached. You basically do some arithmetic to get the percentage. Think of it as a rigorous initial odds-making.

Loads of detail on one such model:

[fivethirtyeight.com]

Once the rating and simulation inputs are settled, they do something quite typical:

Once the adjustments are made, we simulate the regular season 10,000 times to find the average final record of each team and the percentage of simulations that each team makes the playoffs. We use NBA tiebreaking rules to seed teams in the playoffs (including the change this year that makes overall record the top factor in seeding) and then simulate the playoffs 10,000 times to find the winner of the finals.

As with our other sports forecasts, we run our simulations “hot,” meaning that a team’s CARM-Elo rating is updated after each simulated game within a simulated season. This matters more than you might think; essentially, it accounts for the possibility of hot streaks and cold streaks, as well as the increased uncertainty in projecting a team’s fortunes the further you go into the future. This tends to compress playoff and championship odds as compared with running the simulations cold. For instance, as of launch, our model gives the Warriors a 52 percent chance of winning the NBA title, which might sound high — but their probability would be even higher, 73 percent, without this adjustment.

The second paragraph suggests the CHN model's lack of updating on KRACH odds could very well be causing some havoc with their predictions.

Tom Lento
The second paragraph suggests the CHN model's lack of updating on KRACH odds could very well be causing some havoc with their predictions.

As I allude to in our explainer article on the site, I feel like a valid argument can be made to keep KRACH as a snapshot from when the simulation starts. But I can't articulate the reason very well.

On the other hand, I do know that re-calculating KRACH on the fly after every game would be all but impossible. As it is, running 20,000 simulations takes like 4 hours. And each simulation contains a few dozen games or so. At least. If KRACH were re-calculated after each simulated game within each simulation, I think it might take a week to run. Of course, I allow for the fact that I might be doing it wrong.

I could run fewer simulations. The whole thing seems to stabilize at around 3,000 or less. 20,000 is probably overkill. But it does allow for picking up on some outlier possibilities.

adamw

Tom Lento
The second paragraph suggests the CHN model's lack of updating on KRACH odds could very well be causing some havoc with their predictions.

As I allude to in our explainer article on the site, I feel like a valid argument can be made to keep KRACH as a snapshot from when the simulation starts. But I can't articulate the reason very well.

On the other hand, I do know that re-calculating KRACH on the fly after every game would be all but impossible. As it is, running 20,000 simulations takes like 4 hours. And each simulation contains a few dozen games or so. At least. If KRACH were re-calculated after each simulated game within each simulation, I think it might take a week to run. Of course, I allow for the fact that I might be doing it wrong.

I could run fewer simulations. The whole thing seems to stabilize at around 3,000 or less. 20,000 is probably overkill. But it does allow for picking up on some outlier possibilities.

Adam, your CHN model showed a 98% chance before this weekend. Given that we had a 35% KRACH chance of losing on Friday, then 35% of the model's average Cornell NCAA chances after a Friday loss plus 65% of Cornell's average NCAA chances after a Friday win would have to get you to that 98%. So, that would mean, if the model was correctly programmed, that Cornell, after a Friday loss would still have a 94% average chance. So if the 65% after Friday was right, that was quite a huge outlier. And the 94% chance after a loss does not at all pass the smell test. It's not the non-adjusting of KRACH that's causing the problem.

jkahn

adamw

Tom Lento
The second paragraph suggests the CHN model's lack of updating on KRACH odds could very well be causing some havoc with their predictions.

As I allude to in our explainer article on the site, I feel like a valid argument can be made to keep KRACH as a snapshot from when the simulation starts. But I can't articulate the reason very well.

On the other hand, I do know that re-calculating KRACH on the fly after every game would be all but impossible. As it is, running 20,000 simulations takes like 4 hours. And each simulation contains a few dozen games or so. At least. If KRACH were re-calculated after each simulated game within each simulation, I think it might take a week to run. Of course, I allow for the fact that I might be doing it wrong.

I could run fewer simulations. The whole thing seems to stabilize at around 3,000 or less. 20,000 is probably overkill. But it does allow for picking up on some outlier possibilities.

Adam, your CHN model showed a 98% chance before this weekend. Given that we had a 35% KRACH chance of losing on Friday, then 35% of the model's average Cornell NCAA chances after a Friday loss plus 65% of Cornell's average NCAA chances after a Friday win would have to get you to that 98%. So, that would mean, if the model was correctly programmed, that Cornell, after a Friday loss would still have a 94% average chance. So if the 65% after Friday was right, that was quite a huge outlier. And the 94% chance after a loss does not at all pass the smell test. It's not the non-adjusting of KRACH that's causing the problem.

That 65% is dependent on other games. It's not just "if we lose on Friday, we have a 65% chance no matter what.". Had other games gone differently, it could have been higher. Or lower.

These games aren't in a vacuum.

abmarks
How about the simplest explanation- we could just have had a number of PWR comparisons that were razor close? THat matrix uses krach to simulate game results but ultimately fills out the PWR- so the issue is more likely in the PWR than the karch numbers.

Looking at the probability matrix, I think you need to look at the probabilities of landing in any given PWR final position. I'm curious what the distribution was for us across positions before last night, because as of 920 pm, we have crazy high probabilities of landing in 15 or 16, while Penn St. and Providence skew their individual probabilities much more towards the higher PWR finishes

The answer to the riddle will be found by seeing which individual PWR comparisons are likely to get flipped against us as opposed to the other two teams having the same thing happen.

imgur.com

One thing to consider. PWR is no longer a mystery. It's just RPI. If a whole bunch of factors line up perfectly, a team can win a comparison against a higher RPI team. But they need to win head to head against them. At any given time, there are only a few slight differences between the RPI rankings and the pairwise.

Swampy
Two things strangely missing from this rather technical discussion are standard errors and confidence intervals. I'm neither familiar with, nor particularly interested in, the inner workings of the models. But it seems to me that, other things being equal, the standard errors of a prediction regarding a team's prospects will decrease as the team plays more games, but the impact will decrease as the number of games increases. Since Cornell has played fewer games than the factory schools, one would therefore expect an additional game to have a bigger impact on Cornell's standard errors than the competition's. Also, since confidence intervals are wider with smaller N, an additional datum will have a larger impact, percentage wise.

But since such considerations have been absent from the discussion thus far, I have the impression that the predictions are essentially point estimates rather than interval estimates (with some plausibly associated probability distribution) -- the famous "margin of error" routinely misrepresented and misunderstood by the media when reporting survey data.

Comments from someone in the know?

I think the lower number of total games is the best explanation for our volatility.

Talking head analysts like to blather about how good teams win the close ones; the more numerically inclined have known for a while that blowouts are a much better indicator of talent because the close ones are coin-flips, not proof of grit. A team with a high win percentage driven by tons of close wins is probably primed for a fall (but since a season has an arbitrary endpoint, it may not happen until after the banner is hung). In baseball it probably means being propped up by a solid bullpen, in hockey, by the goalie.

There are flaws in using goal differential, certainly, and it probably isn't the easiest thing to properly integrate, but a system that doesn't integrate it at all will always overstate the likelihood of success of teams like Cornell.

We can and should win this afternoon, though, so let's do that. Whatever our team is, it spent a season showing that we're modestly better than Clarkson, so let's get that result on the board.

jkahn
Adam, your CHN model showed a 98% chance before this weekend. Given that we had a 35% KRACH chance of losing on Friday, then 35% of the model's average Cornell NCAA chances after a Friday loss plus 65% of Cornell's average NCAA chances after a Friday win would have to get you to that 98%. So, that would mean, if the model was correctly programmed, that Cornell, after a Friday loss would still have a 94% average chance. So if the 65% after Friday was right, that was quite a huge outlier. And the 94% chance after a loss does not at all pass the smell test. It's not the non-adjusting of KRACH that's causing the problem.

It's showing 88% now.

I'm 99.999% sure it's not programmed incorrectly. If it were, a lot of other teams would have crazy effects too. That is not the case. This big swing only happened to Cornell.

Cornell apparently has been slow to react to the forechecks some teams have used against them. I hope this is the case rather than a lack of talent which limited them to three shots in two periods of hockey Friday. It's easier to change your clear and your attack than it is to change your personnel.

9

Hooking
Cornell apparently has been slow to react to the forechecks some teams have used against them. I hope this is the case rather than a lack of talent which limited them to three shots in two periods of hockey Friday. It's easier to change your clear and your attack than it is to change your personnel.

And besides, it's past the trading deadline.

ugarte
We can and should win this afternoon, though, so let's do that. Whatever our team is, it spent a season showing that we're modestly better than Clarkson, so let's get that result on the board.

This. Me want Lake Placid.

Trotsky

ugarte
We can and should win this afternoon, though, so let's do that. Whatever our team is, it spent a season showing that we're modestly better than Clarkson, so let's get that result on the board.

This. Me want Lake Placid Chicago.

FYP.

In case anyone's curious about out of conference rooting interests, we have a conundrum in UNH vs UML.

A UNH win helps our RPI, but we absolutely don't want UNH winning that conference tournament. This is only the quarterfinals, so I think I'm tentatively rooting for UNH. For what it's worth, we probably want UML or BU (ugh) to win the tourney itself. Notre Dame is a couple spots ahead of us and passable, and BC is right on the bubble. If BC wins it, they probably rise enough that they are no longer on the bubble, so that's probably less of a concern than it looks.

There's a WCHA game that doesn't appear to matter at all (winner of Minnesota State and Michigan Tech plays Bowling Green in the finals), but I'm sure has some RPI implications for someone somewhere.

In the NCHC, Omaha plays Western Michigan. Barring some unforeseen RPI implications, we want Western Michigan, and then we want Western Michigan, Denver, or UMD to win that tournament, instead of North Dakota, who we have a shot of passing. Again, ND winning the tournament puts them high enough that it's probably not as much of a concern, but I live in South Dakota (which is the better Dakota) and North Dakota sucks, so...

The Big10 tournament is next week. We want Minnesota to win it. Beyond that, it'd be swell if Ohio State and Penn State could lose in the 1st round, as they are in the two spots right below us. Wisconsin is the 2 seed with a bye in the 6 team tourney, but they're below the bubble. So really, go Minnesota.

In the AH, Air Force is sitting in a tie for 15th. Hard to believe they could wind up above the bubble, even if they win the AH, but that's the best possibility. They face the winner of Army and Mercyhurst next weekend in the semis.

Point being, Let's Go Red.

Final Lynah game for the seniors; hope for some extra inspiration.

Dafatone
In case anyone's curious about out of conference rooting interests, we have a conundrum in UNH vs UML.

6 first period goals by Lowell have rendered this conundrum irrelevant.

ugarte

Dafatone
In case anyone's curious about out of conference rooting interests, we have a conundrum in UNH vs UML.

6 first period goals by Lowell have rendered this conundrum irrelevant.

Whoa! If Umile is indeed going out the door, not exactly the way you'd want to go.

Dafatone
One thing to consider. PWR is no longer a mystery. It's just RPI. If a whole bunch of factors line up perfectly, a team can win a comparison against a higher RPI team. But they need to win head to head against them. At any given time, there are only a few slight differences between the RPI rankings and the pairwise.

The only odd-ball situation existed before this weekend, as Providence sat two spots behind us at 12 despite having a higher RPI than both us (at 10) and Notre Dame (at 11). While PC won our comparison, they lost the H2H comparison with both Notre Dame AND BC. That meant that we won more comparisons than PC, and the same # as Notre Dame, but because we had a higher RPI than ND, we got the higher "ranking."

That has untangled a bit now, as ND has the higher RPI, so we've dropped that comparison and are now tied in comparisons with PC, but we lose the RPI and get the lower "ranking."

Since PC is out of the HEA playoffs, their RPI won't change dramatically from here on. We sit (before the conclusion of Sunday's game) 0.0007 behind them.

will 1-1 be enough next weekend.

2-0 this weekend puts us at 8 and a loss wouldnt have hurt. 2-1 put us in first team out territory again,

upprdeck
will 1-1 be enough next weekend.

2-0 this weekend puts us at 8 and a loss wouldnt have hurt. 2-1 put us in first team out territory again,

1-1 will most certainly be enough. The question is whether 0-1 is enough. I'd say most likely, but not certainly.

I have a hard time seeing 0-1 being good enough with the teams that are likely to lose already being ahead of us.

upprdeck
I have a hard time seeing 0-1 being good enough with the teams that are likely to lose already being ahead of us.

It really depends on who wins the tournaments. It looks like we are unlikely (not impossible. Unlikely) to fall below 14th. That could get in, or not.

if osu/psu lose it would help us a would BC. i dont know if another lose drops us below VT but i doubt it.

that would leave Wisc as the one to be worried about.

just glad we survived this weekend so we have something to talk about.

is the pwr what if tool around this year?

playing some with the college hockey pwr matrix. if i plug in us losing / nd losing and psu or osu losing we actually go up most of the time. thats good to see.

even psu/osu winning with nd losing keeps us at 14.. so we wont need crazy stuff just a couple good results. just 1 of the 3 losing might be enough

My idea is, we win and not find out.

And QU beating SLU moves Q to 20, gives us 0.0001 QWB points, and moves us over NoDak.

Jeff Hopkins '82
And QU beating SLU moves Q to 20, give us 0.0001 QWB points, and moves us over NoDak.

And somehow lifts the Irish by .0001, as well.

now up to 9th. ND trails by .0001

You Are The Committee is now up! [www.collegehockeynews.com]

I threw together all the worst results I could think of and had us in 14th with the top 13 getting in. We missed by like .0003 RPI or something.

We look likely (likely!) to be in.

Thanks for link to YATC - was coming here to plug it so you can all see for yourself. Please report any glitches.

I'm going to re-run the Probability Matrix now too ... so you can start seeing that as the iterations tally up.

The Matrix has already found a way for Cornell to not get in .... So you'll have to play with YATC's 3 million-plus permutations to figure out what that scenario is.

[www.collegehockeynews.com]

well the question is will mich/MSU or Quin take away a bid that could drop it to 12 and there a ton of ways to not get in.

but if you play the bad things happen to cornell game you can have us lose and osu/psu/NDak all get to the finals and then the only thing that looks bad is bc/nd with bc winning.

it takes almost all 4 of the bad things to happen to knock us out.. if you have PSU/Ndak/BC winning and osu losing early we still get in.

adamw
The Matrix has already found a way for Cornell to not get in .... So you'll have to play with YATC's 3 million-plus permutations to figure out what that scenario is.

[www.collegehockeynews.com]

Already found it. We lose to Union...Quinny wins ECAC over Union, BC wins HE over ND, Ohio State beats Penn State for the Big 10, NoDak wins NCHC over Western. We finish 15th with three autobids below us.

God help me, but I think I have to root for Harvard in oder to keep Quinny from getting the autobid.

Edit: It doesn't matter who wins the ECAC. But we need BC and tOSU to lose.

There's a bunch of scenarios where Cornell doesn't get in among the 3 million ... but they all have minuscule odds, which is why the Matrix has their chances at 99%

Jeff Hopkins '82

adamw
The Matrix has already found a way for Cornell to not get in .... So you'll have to play with YATC's 3 million-plus permutations to figure out what that scenario is.

[www.collegehockeynews.com]

Already found it. We lose to Union...Quinny wins ECAC over Union, BC wins HE over ND, Ohio State beats Penn State for the Big 10, NoDak wins NCHC over Western. We finish 15th with three autobids below us.

God help me, but I think I have to root for Harvard in oder to keep Quinny from getting the autobid.

Edit: It doesn't matter who wins the ECAC. But we need BC and tOSU to lose.

We also don't want Wisconsin to win the Big10.

with us back up to 9 it really is HE controlling most everything bad. if bc/nd lose we stay in as long as bid stealers dont show up all over.

Lots of talk about how likely we are to be "in." But how about some talk about how likely we are to finish 8th (or higher) and be a two-seed? I'm guessing that would take two wins in Lake Placid, but would that be enough to insure it? Is it even possible with just one win?

andyw2100
Lots of talk about how likely we are to be "in." But how about some talk about how likely we are to finish 8th (or higher) and be a two-seed? I'm guessing that would take two wins in Lake Placid, but would that be enough to insure it? Is it even possible with just one win?

Play with You Are the Committee ... It will answer all your wildest dreams.

YATC: chalk except we win out puts us in PWR 8.

PlayoffStatus.com odds:

.99 Make the NCAA
.47 QF
.19 F4
.09 F
.04 National Champion

Trotsky
YATC: chalk except we win out puts us in PWR 8.

Careful what you wish for. That could mean playing 3 seed North Dakota in Fargo - or 3 seed Providence in Providence.

Dafatone

jkahn

adamw

Tom Lento
The second paragraph suggests the CHN model's lack of updating on KRACH odds could very well be causing some havoc with their predictions.

As I allude to in our explainer article on the site, I feel like a valid argument can be made to keep KRACH as a snapshot from when the simulation starts. But I can't articulate the reason very well.

On the other hand, I do know that re-calculating KRACH on the fly after every game would be all but impossible. As it is, running 20,000 simulations takes like 4 hours. And each simulation contains a few dozen games or so. At least. If KRACH were re-calculated after each simulated game within each simulation, I think it might take a week to run. Of course, I allow for the fact that I might be doing it wrong.

I could run fewer simulations. The whole thing seems to stabilize at around 3,000 or less. 20,000 is probably overkill. But it does allow for picking up on some outlier possibilities.

Adam, your CHN model showed a 98% chance before this weekend. Given that we had a 35% KRACH chance of losing on Friday, then 35% of the model's average Cornell NCAA chances after a Friday loss plus 65% of Cornell's average NCAA chances after a Friday win would have to get you to that 98%. So, that would mean, if the model was correctly programmed, that Cornell, after a Friday loss would still have a 94% average chance. So if the 65% after Friday was right, that was quite a huge outlier. And the 94% chance after a loss does not at all pass the smell test. It's not the non-adjusting of KRACH that's causing the problem.

That 65% is dependent on other games. It's not just "if we lose on Friday, we have a 65% chance no matter what.". Had other games gone differently, it could have been higher. Or lower.

These games aren't in a vacuum.

I clearly understand that. That's way I said that the 65% would be an "quite a huge outlier" if the 98% prior to Friday were correct. Fortunately, we've now taken away the the pre-Friday 12% chance of getting swept by Clarkson, and 27.5% chance of losing that series, so we don't have to worry about those negative scenarios.

adamw

Trotsky
YATC: chalk except we win out puts us in PWR 8.

Careful what you wish for. That could mean playing 3 seed North Dakota in Fargo - or 3 seed Providence in Providence.

I actually don't think playing the Friars there is all that bad. It's not like Providence has a massive draw and there would be a lot of Cornell supporters there anyway.

Iceberg

adamw

Trotsky
YATC: chalk except we win out puts us in PWR 8.

Careful what you wish for. That could mean playing 3 seed North Dakota in Fargo - or 3 seed Providence in Providence.

I actually don't think playing the Friars there is all that bad. It's not like Providence has a massive draw and there would be a lot of Cornell supporters there anyway.

That was my argument against putting them there in 2015 ... I said the Committee shouldn't bother because other Eastern teams will draw just as well as them anyway. Turns out, Providence did have a massive draw there, so I was wrong.

adamw

Iceberg

adamw

Trotsky
YATC: chalk except we win out puts us in PWR 8.

Careful what you wish for. That could mean playing 3 seed North Dakota in Fargo - or 3 seed Providence in Providence.

I actually don't think playing the Friars there is all that bad. It's not like Providence has a massive draw and there would be a lot of Cornell supporters there anyway.

That was my argument against putting them there in 2015 ... I said the Committee shouldn't bother because other Eastern teams will draw just as well as them anyway. Turns out, Providence did have a massive draw there, so I was wrong.

Tell you what, if it means we win the ECAC & go away, I'll take it.

I'd like to win the NCAA, but the chances are small & winning the ECAC would be great for our team, fans & recruiting.

Pairwise predictor might keep me up the next four nights

wakester2468
Pairwise predictor might keep me up the next four nights

So long as you use the right one, you're good ... the original, invented by a Cornell alum :

[www.collegehockeynews.com]

Bracketology Blog College Hockey:

What I Believe – Monday Edition
By Jayson Moy • Senior Writer • March 13, 2017

Here’s what I believe after this weekend’s action and by playing around with the PairWise Predictor.

In:

Big 10 – Minnesota
ECAC – Harvard, Union
Hockey East – Lowell, BU
NCHC – Denver, UMD, Western Michigan
WCHA – Winner of Bowling Green/Michigan Tech

That makes nine teams in, leaving seven spots left.

There are 15 teams for those seven spots

Atlantic Hockey – Canisius, Air Force, Army, Robert Morris
Big 10 – Wisconsin, Ohio State, Penn State, Michigan, Michigan State
ECAC – Cornell, Quinnipiac
Hockey East – BC, Notre Dame, Providence
NCHC – North Dakota

Teams that need to win its Championship in order to get in (seven):

Atlantic Hockey – Canisius, Army, Robert Morris
Big 10 – Wisconsin, Michigan, Michigan State
ECAC – Quinnipiac

Teams that can still get in at-large (eight):

Atlantic Hockey – Air Force (How? All top seeds win, except that one of Penn State or Ohio State must lose on Thursday and Air Force must win on Friday)
Big 10 – Ohio State, Penn State
ECAC – Cornell
Hockey East – BC, Notre Dame, Providence
NCHC – North Dakota 

Read more: [www.uscho.com]

Last before the league championships.

East Regional (Providence):

13 Penn State vs. 4 Western Michigan
12 Providence vs. 7 Union

Northeast Regional (Manchester):

14 Ohio State vs. 3 Harvard
9 Cornell vs. 6 Boston University

Midwest Regional (Cincinnati):

15 Canisius vs. 2 Minnesota-Duluth
10 Notre Dame vs. 5 Minnesota

West Regional (Fargo):

16 Michigan Tech vs. 1 Denver
11 North Dakota vs. 8 Massachusetts-Lowell

Conference breakdowns

NCHC — 4
Hockey East — 4
ECAC Hockey — 3
Big Ten — 3
WCHA — 1
Atlantic Hockey – 1

Movement

In: Michigan Tech, Ohio State
Out: Bemidji State, Wisconsin

Read more: [www.uscho.com]

So here we are again in the undesirable position of cheering for Harvard on Friday night but different than other times, we will have to do it in person. The only saving grace is
it's Quinnipiac who ranks right up there with teams to root against. The last thing we want to see is Q to get the automatic qualifier for the ECAC making it a little merky for the Big Red should they lose to Union.

To be honest, I'd root for Harvard against several ECAC opponents. Q and Union, certainly.