2023 ECAC Post Season

[andyw2100]

There are 3 finals that matter tomorrow - CCHA, ECAC and NCHC, unless the 3 underdogs win all three, we are in NCAA

When was the last time the ECAC had four teams in the tourney?

[BearLover]

So we need any one of:

St. Cloud scores in OT and beats CC
Harvard over Colgate - Hey Trotsky where ya at on this?
BU over whoever wins UML/Merrimack

Right?

Right. Harvard and BU will be the favorites, so we have a pretty good shot.

The 99.7% number from the CHN model was a complete joke. I'll keep harping on it until people stop citing it. It was cited by the Cornell Daily Sun, the ELynah twitter account, the popular "Everything College Hockey" social media account, and the literal commentators of tonight's game. Publishing this model is doing everyone a disservice. It isn't calibrated to reality. It isn't backwards-tested. No one made any attempt to check whether weighting by KRACH yields accurate predictions. But they publish the model anyway and people keep referring to it.

I have thoughts on the probabilities, but I'm gonna wait til after tomorrow to share them. Wrath of the whatever from high atop the thing and all.

I am not gonna wait.  Even if all three games that matter were just toss-ups, the odds of all three going wrong would be .5*.5*.5 = .125 = 12.5%
If you give favorites a 60% chance to win, then the odds of three upsets would be .4 * .4 * .4 = .064 = 6.4%

So, even after all the bad results today, the odds that Cornell makes the tournament are still in the 93.6% to 87.5% range.

I know it might not feel right, but trust the math.

Throwing your hands up and saying that it must be wrong makes me think that you are the guy who loses the bet, not believing that in a room of 40 people, the odds are overwhelming that two have the same birthday. This also feels like it can't possibly be right, but it is.

Huh? I agree with your odds (for the most part—I think weighting for the favorites is closer than 60/40). I don't agree with the 99.7% number ever being correct, though. The 99.7% number wasn't based on 60/40 weighting for favorites. Rather, it was based on completely absurd weighting based on KRACH, which gave Q something like a 90% chance of beating Colgate.

To add to this, here are the probabilities based on "weighting by KRACH":
SCSU has a 75% chance of beating CC
BU has a 59% chance of beating Merrimack
Harvard has a 75% chance of beating Colgate

The chances of all three of those things failing is 2.6%, quite a bit lower than the 6.4% or 12.5% from
nshapiro's numbers.

Again, we see here that, when "weighting by KRACH," the further apart two teams' KRACH scores, the more extreme(ly wrong) the probabilities are. As I have previously mentioned, according to this method, Minnesota would beat Dartmouth 97% of the time.

Here's the thing, though—this discussion shouldn't even be happening, because the burden on proof isn't on random laypeople to disprove the methodology. Rather, the burden of proof is on the people publishing these models to show how and why weighting by KRACH yields predictive outcomes. To my knowledge, no one has put forth any argument why these models should be trusted. People just pick whatever random inputs they want, publish a model, and then everyone else eats it right up.

[nshapiro]

Maybe I am getting too tired, but YATC is now showing Merrimack ahead of us even if they lose, so we might need one of two favorites to win, Hahvahd or St. Cloud, not one of three.

[BearLover]

Maybe I am getting too tired, but YATC is now showing Merrimack ahead of us even if they lose, so we might need one of two favorites to win, Hahvahd or St. Cloud, not one of three.

I'm getting Cornell ahead of them when I do it

[ugarte]

Maybe I am getting too tired, but YATC is now showing Merrimack ahead of us even if they lose, so we might need one of two favorites to win, Hahvahd or St. Cloud, not one of three.

I'm getting Cornell ahead of them when I do it

me too. every possible combination of underdogs except BU and we're still in on USCHO's doohickey.

[kingpin248]

When was the last time the ECAC had four teams in the tourney?

Four years ago; Cornell and Quinnipiac lost in the second round, Harvard and Clarkson in the first.

[andyw2100]

When was the last time the ECAC had four teams in the tourney?

Four years ago; Cornell and Quinnipiac lost in the second round, Harvard and Clarkson in the first.

Thanks!

[jtwcornell91]

How does a playoff overtime impact pwr? Half a win?

I thought 5x5 OT results counted the same as regulation.  Do they not?

[jtwcornell91]

No one made any attempt to check whether weighting by KRACH yields accurate predictions.

Yes, we have: https://arxiv.org/abs/2001.04226

[jtwcornell91]

No one made any attempt to check whether weighting by KRACH yields accurate predictions.

Yes, we have: https://arxiv.org/abs/2001.04226

Figure 8 in particular is related to a previous eLynah discussion about whether probabilities too close to 1 are inaccurate.

[upprdeck]

Bu winning helps in a couple ways.

It gets us in and it stops Merrimack from jumping over us which would keep us from playing a #1 seed someplace potentially,

N Mich helps us there too. All the Favs and NMich and we have 4 teams behind us n the PW before they start moving teams around.

But if that all happens there will be a bunch of moving any way

[djk26]

Ugh--so this weekend is the double torture of losing to Harvard--and then having to root for them in the next game.::pain::

[BearLover]

No one made any attempt to check whether weighting by KRACH yields accurate predictions.

Yes, we have: https://arxiv.org/abs/2001.04226

Wow, this is amazing. I had no idea! You are one of the authors? I'm going to try my best to read this as a non-math major. (Though feel free to offer the cliff notes if you'd like.)

[scoop85]

BU's top center Skoog is suspended tonight due to a boarding penalty last night

[andyw2100]

Ugh--so this weekend is the double torture of losing to Harvard--and then having to root for them in the next game.::pain::

Root for them to tie us in number of ECAC titles? Sorry-can't do it.

I'll be there rooting for Colgate and hoping that BU or Saint Cloud get it done.