Everyone’s favorite hockey ranking scheme, KRACH, can be used to predict likelihoods of game outcomes. So, without further ado, here’s what KRACH says the likelihood of each team winning each round is.
The Midwest region is most balanced, but the brutal schedules actually make it less likely (22.7%) that the champ will emerge from that region than the lopsided East (28%).
The most lopsided is the East, with QU favored by KRACH to have an 84% chance of beating Canisius. What a difference seeding makes — QU is given an 18.7% chance of winning it all vs. Minnesota’s 14.9%, despite Minnesota just eking out the KRACH edge over Quinnipiac.
Both #2 New Hampshire and #2 Miami are actually underdogs to their WCHA 3-seed foes, Denver and MSU-Mankato, respectively.
| KRACH | West | Game 1 | Game 2 (Region Champ) | Game 3 (Frozen four semifinal) | Game 4 (National Champ) |
| 157.487 | 1. Minnesota | 65.64% | 42.99% | 25.75% | 14.86% |
| 82.4324 | 4. Yale | 34.36% | 17.21% | 7.56% | 3.17% |
| 100 | 2. North Dakota | 63.02% | 27.94% | 13.61% | 6.34% |
| 58.6682 | 3. Niagara | 36.98% | 11.87% | 4.25% | 1.45% |
| Northeast | |||||
| 127.142 | 1. Mass.-Lowell | 56.72% | 32.35% | 17.27% | 9.07% |
| 97.0033 | 4. Wisconsin | 43.28% | 21.77% | 10.20% | 4.68% |
| 92.3517 | 2. New Hampshire | 48.24% | 21.69% | 9.90% | 4.42% |
| 99.1066 | 3. Denver | 51.76% | 24.18% | 11.45% | 5.31% |
| East | |||||
| 157.423 | 1. Quinnipiac | 83.83% | 53.06% | 31.97% | 18.67% |
| 30.3748 | 4. Canisius | 16.17% | 4.05% | 0.92% | 0.20% |
| 100.909 | 2. Boston College | 55.73% | 25.17% | 12.41% | 5.90% |
| 80.1731 | 3. Union | 44.27% | 17.72% | 7.72% | 3.24% |
| Midwest | |||||
| 111.145 | 1. Notre Dame | 55.05% | 28.24% | 13.73% | 6.85% |
| 90.736 | 4. St. Cloud St | 44.95% | 20.78% | 9.09% | 4.09% |
| 105.133 | 2. Miami (OH) | 49.82% | 25.35% | 11.99% | 5.82% |
| 105.899 | 3. Minnesota St | 50.18% | 25.63% | 12.16% | 5.93% |
How accurate has this analysis been in predicting the NCAA winner in years past?
Well, the problem is that it’s not a repeated game. If KRACH predicts something has a 75% chance of occurring and it does (or doesn’t), was the prediction right or not?
I’ve been pondering some ways to test KRACH’s predictive abilities (e.g. seeing over time how many of the 70-80% games were won by the favored team, and over dozens of games that should settle into the 70s).
As I noted in my previous post, I think it’s predictive value declines significantly in tournaments, and there’s not a lot of data. Even with five years of data, I only have 75 games total, so we’re not going to get a lot in each band.
It would be interesting to see a prediction model based on goaltending stats. Seems like reliable goaltending is half the game.