I rankled some Michigan Tech fans a week or so back when I intimated that they would have a very hard time catching Bemidji State. I still think that’s true. Here is my case.
- Through 12 WCHA games, Bemidji State is 11-1-0-0 for 35 league points.
- Through 10 WCHA games, Michigan Tech is 6-1-1-2 for 21 league points.
- With 18 games remaining, Tech can finish with a maximum of 75 league points.
- Tech and Bemidji still play four times this season; in this scenario, Tech would win the A tiebreaker 4-0-0.
- With four losses, Bemidji cannot win enough league games to best Tech; the Beavers would max out at 12-0-0-4 for 73 league points.
Is Tech running the table a ridiculous enterprise? No, but it is an unlikely one, starting with the fact that they have to beat an unbeaten team four times.
With that said, I want to point out something about point #5 above: Bemidji would be finishing with .750 hockey down the stretch; as of Week 8, BELOW expects BSU to beat any average WCHA team 77% of the time. So even the elite scenario for the Huskies — winning out, sweeping the Beavers — only just barely gets them past the post.
But you deal in probabilities!
Yes, I do. Let’s talk about one.
Does BELOW directly deal in things like injuries and suspensions? No, it doesn’t, but let’s go with this ridiculous example:
- Michael Bitzer is abducted by aliens and the entire Bemidji goaltending cadre is left with noodly-armed Muppets. They’re going to get blown out, right?
- Tech is blown out both games in Alaska and finishes with a 1593 BELOW, causing Mitch’s Misfits to show up at my house with flaming pitchforks.
- Tech levels the Hoover-goalies with two blowout wins. Tech BELOW 1666, BSU BELOW 1640.
- Lake Superior, which still does have a solid scoring offense, polishes their peashooters and also destroys the Beavers, who are now 11-1-0-4, still with 35 points, and still in 1st place … with a BELOW of 1558.
At 1558, BSU would be behind Tech, Minnesota State, and Bowling Green in BELOW (in some order). Instead of being expected to win 77% of the time, it would be expected to win just 58% of the time — and that’s after losing any chance at the A tiebreaker with Tech.
Is this scenario likely? I don’t believe in aliens, but it’s entirely possible that Bitzer could be injured during practice and be out for the season. It’s possible that his backups could all be terrible, although Reid Mimmack and Jesse Wilkins have a solid track record. It’s also possible that all three Fitzgeralds could be injured along with any of the seniors leading the Beavers in scoring.
It’s going to take something like that for the Beavers to just fall off the table. The above scenario requires Tech to get blown out by Fairbanks this weekend, thus lowering their BELOW and maximizing their gains with a sweep of Bemidji. It’s … far-fetched.
But back to probabilities
ABOVE calculates the probabilities of these sorts of events happening — not abductions in specific, but changes in valuation of the team in general. BELOW is an Elo-style rating, so all you know from the number is that the system thinks that a team is only so good relative to the average squad (1500). The probability of a blowout in either direction, however small, is non-zero, and those blowouts do have a big impact. When the needle moves, the probabilities later in the season can change significantly.
The crazy scenario above has Bemidji losing its next four league games. BELOW says, on average, that the Beavers would lose four games in their final 16. But losing the next four games, especially in spectacular fashion, changes BELOW’s opinion of the Beavers significantly. Once that changes, ABOVE is going to assign different probabilities to the Beavers, and that will be reflected in the results.
Let’s look at the opposite scenario: Tech blows out Fairbanks (Tech BELOW 1704), Bemidji blows out Tech (BSU 1766, Tech 1651), Bemidji blows out Lake (BSU 1783). Instead of being -155 in BELOW, BSU would be +70. That’s because BELOW already thinks that Bemidji is white hot.
Lather, Rinse, Repeat
A Monte Carlo simulation is the way to figure these things out, and that’s what I use. I estimate the chances something could happen, I roll a die, I re-calculate, and I roll again. I’ve only use this technique within a series — simulation 1 says the home team routs on Friday, simulation 2 says the away team wins in a shootout — but it’s not difficult to extend the concept forward.
If we look at the next two weeks, all four teams that play in Week 9 also play in Week 10. I could estimate the probabilities for the Week 10 games now, but I won’t know what we’ll know after Week 9. But I can come up with a range of probabilities for the Week 9 games and input them into a simulation of Week 10 … and then forward through the rest of the season.
(If you’re curious, Bemidji would have to keep its undefeated run going through its trip to BG to go above 1800. It’s harder to move farther from the mean the farther from it you start.)
Has Bemidji sewn this up? From an account perspective, they haven’t, but it’s getting closer. Forced to give a guess, I’d say that it’s about 80% likely that Bemidji State has already won the MacNaughton Cup.
I hate everything.