2015 NCHC Mid-Season Check-In

In this photo, my predictions are Denver’s Joey LaLeggia and the season is North Dakota’s Luke Johnson. 


Mid-season evaluation time! Seeing as how there are no more NCHC games until January 9, let’s look at how the model is holding up. Technically, I said last month the model won’t make reliable predictions until mid-January, and that is still true. It would be better to look at the model after the 14-game mark. But every other NCAA hockey commentator is jumping off the bridge currently, so I will too.

As a reminder, I made the following predictions in September before the season started:

2015 NCHC Prediction

Team '15 Shot% '15 Save%  '15 Poss% '15 Points* '15 Finish
North Dakota 11.26%  92.03% 49.50% 48 1
Nebraska Omaha 10.82% 90.45% 49.50% 41 2
Minnesota Duluth 9.05% 89.35% 54.50% 36 3
St. Cloud State 10.02% 90.21% 49.00% 36 3
Denver 9.44% 91.45% 47.50% 36 3
Miami 7.81% 90.90% 52.75% 34 6
Western Michigan 8.68% 90.45% 48.50% 30 7
Colorado College 8.31% 89.72% 48.75% 27 8
ALL NCHC 9.43% 90.57% 50.00% 288
*All 14-15 expected points +-4.85

That was based on analysis of last year’s performances plus imputed values for roster changes. Now we have some real game data – at least seven games’ worth for each team. So, how close was the model in prediction shot-based metrics, and thus year-end finishes? You can see here how the model has adjusted week by week, and what the current predictions are. It’s starting to line up with the standings, which is a good sign, but it that by chance or by design? Let’s take a look, and let’s talk about what to expect for the rest of the season:

hlogo-CC  Colorado College

Date Shot% Save% Poss% Points Finish
Preseason 8.31% 89.72% 48.75% 27 8
Mid-season 5.85% 87.95% 38.2% 0 8

Colorado College is struggling on every level. Looking forward, it only gets tougher. They have one of the toughest remaining league schedules, if only by virtue of the fact that they don’t have to play themselves. And even if they did, I don’t know that they’d earn many points. My model still has them earning zero points, so with the point they won in a shootout loss, I guess you could say they’re beating expectations?

I think the real questions is -why so much worse than expected? There were a few roster shake-ups, but not that many, on CC or on other league teams. You have to look to coaching on this one – adjusting to a new system takes time and new recruits. Certainly the strength of every – literally every – other NCHC team plays a part, too, though. Will the Tigers get more than one point? Yeah, probably. But I’m thinking single digits.


hlogo_DEN  Denver

Date Shot% Save% Poss% Points Finish
Preseason 9.44% 91.45% 47.50% 36  3t
Mid-season 11.26% 89.77% 50.8% 42 5

Flashes of brilliance from Denver coupled with a few real dud performances have them in the middle of the pack currently. They are shooting much better than expected, thanks to some veterans who have stepped up their play. I expected goaltending to decrease only slightly, but it turns out the Pios are missing Sam Brittain much more than I thought they would. Nevertheless, it’s clear I underestimated Denver’s ability to control play – they’re a better-than-average possession team, far from the league-worst they were last year.

Denver could potentially get 42 points like the model predicts, but I wouldn’t count on it just yet. Keep in mind they’ve played only seven NCHC games – less than any other team – so these predicitions are the least sound at this point. They also have a tough remaining schedule, considering they have to play 10 of the remaining 17 games against UND, UMD, Miami and UNO.


hlogo_MIA Miami

Date Shot% Save% Poss% Points Finish
Preseason 7.81% 90.90% 52.75% 34  6
Mid-season 9.88% 91.34% 57.5% 52 1

So, yeah, I was probably wrong about Miami. I knew their possession would be high, but not this high. And given last year’s numbers, I didn’t expect the shooting to get so much better. Yeah, yeah, regression toward the mean – I get that. But this is arguably over-correction that goes beyond a statistical effect, so kudos to Coach Blasi for turning this team around in the offseason.

The model has Miami finishing first with 52 points. That’s a crap-ton of points, equivalent to 17 wins and a shootout loss. Miami would need to win 10 of their last 14 to get there. Given how they’ve been splitting with top teams, I doubt that happens. But, Miami also has a relatively easy schedule going forward. Don’t bet on 52 points, but don’t rule it out either. And don’t bet against Miami fighting for first place when North Dakota comes to town on March 6.


hlogo_UMD Minnesota-Duluth

Date Shot% Save% Poss% Points Finish
Preseason 9.05% 89.35% 54.5% 36  3t
Mid-season 9.46% 91.42% 56.6% 48  2t

Duluth is the team most living up to the model’s expectations. Shooting is right on target, goaltending is a bit better than expected, and possession is about what we expected. Nevertheless, they are slightly better than expected on all of those metrics, so the formula is expecting 48 points from them at this point. Credit Toninato and Kaskisuo. I think that’s going to fluctuate over the next couple of series, but which direction is up to Minnesota-Duluth.

Going forward? On the road against North Dakota, Denver and Miami. I don’t see any sure things for Duluth for the rest of the NCHC schedule, but they’re all winnable. If they split every remaining series, they end up with 42 points. I think they’ll do better than that, so I like the 48 point prediction. Is that good for 2nd? We’ll see.


hlogo_UNO Nebraska-Omaha

Date Shot% Save% Poss% Points Finish
Preseason 10.82% 90.45% 49.5% 41  2
Mid-season 12.60% 92.12% 44.3% 48  2t

Who is this team? They’re defying expectations in a lot of ways. They have the most accurate shooting in the NCHC, but they’re taking very few shots (related?). Massa’s been a hero in net, but can he keep it up, and how many times this season is he going to have to stop 40+ shots? Weirdly, in the four games he’s face 39 or more shots, UNO is 4-0. This team is single-handedly breaking this model.

In 2015, UNO either continues this improbable combination of shot%, save% and possession, and finishes with 48 points, or they regress hard like the last three years. The model says the former, but the model is designed to say the former. This is why we play the games, of course. But look, UNO’s remaining schedule is manageable. Toughest series? Probably Duluth on the road. Maybe UND at home. I would imagine they’ll be favored in nearly every other game. Ten more wins and 30 more points? Believe it, Mav fans, it’s possible.


hlogo_UND North Dakota

Date Shot% Save% Poss% Points Finish
Preseason 11.26% 92.03% 49.5% 48  1
Mid-season 9.83% 93.07% 49.3% 45  4

The model nailed North Dakota – except for one tiny detail. Before the season, I thought UND would run away with the regular season title. That was premature. It’s not that UND has performed any different than expected (actually, they’ve performed exactly as expected). It’s that other teams have stepped up their game from last year, so assuming CC is as bad as they look, 45 or even 48 points probably won’t be enough to win the NCHC.

It won’t be an easy road for Hakstol & Co., either. Duluth comes to town after break, then trips to Omaha and Oxford later. Fourth place seems really low for this team given pre-season expectations, but it’s a stacked year, and the injuries are hurting. North Dakota is know to turn up the volume after January 1 though, so don’t ever – ever – count them out of the race.


hlogo_SCSU St. Cloud State

Date Shot% Save% Poss% Points Finish
Preseason 10.02% 90.21% 49.0% 36  3t
Mid-season 7.47% 90.35% 51.4% 28  6

If you can’t shoot, you can’t win hockey games. Simple as that for SCSU at this point in the season. Don’t get me wrong, six of their eight games have been against the league’s three best goaltenders, but even then you’d expect the shooting to be closer to 8-8.25%. Otherwise, about what was expected. Before the season, I gave SCSU an outside shot at home ice. As 2015 begins, consider that shot way, way outside.

Going forward, the Huskies’ schedule looks pretty brutal, I’m not gonna lie. The good news? Only one series this season against Miami. The bad news? Only one series this season against Colorado College. Of their 16 remaining games, four are against teams who should finish below them, and 12 are against teams ranked in the top 12 nationally, including road series in Duluth and Grand Forks. The model says SCSU earns 28 points. Are there 21 points left on the schedule? I mean, technically…


 hlog_WMU Western Michigan

Date Shot% Save% Poss% Points Finish
Preseason 8.68% 90.45% 48.5% 30 7
Mid-season 7.83% 88.85% 51.1% 25 7

I thought earlier this season that the Broncos were a good team, but not good enough to make waves in the NCHC. I stand by that. They did, after all, score 8 goals in a win against the defending national champions. They’d be the third- or fourth-best team in the WCHA, B1G or ECAC. Nevertheless, they’re performing a bit worse than I expected, again because the competition is so tough. Great possession though? Who’d have guessed? (The answer is not me).

Up ahead? Western has, without a doubt, the toughest schedule of any NCHC team. With 14 games to go, they have four against Duluth, and two each against Miami, Omaha, and North Dakota. Their “easy” games – Colorado College and St. Cloud – are all on the road. The model thinks WMU earns 25 points. This means they need 15 more. That’s five more wins. Find me five more wins on their schedule.


Final Notes

In the interest of transparency, let’s examine my record. There were 24 shot-based data points to predict. Let’s consider a shot% and save% within 1% highly accurate, and a possession within 2.5% accurate. Given those stipulations, I am 8-16. Not too hot, I understand.

But, to humor me, take a look at the trends. I’ve indicated whether my method predicted each stat to be up or down from last year (UP/DW), and whether that is correct (Y/N):

Team '15 Shot% '15 Save%  '15 Poss%
North Dakota UP-N  UP-Y UP-Y
Nebraska Omaha UP-Y UP-Y  DW-Y
Minnesota Duluth UP-Y DW-N  UP-Y
St. Cloud State DW-Y DW-Y  DW-N
Denver UP-Y DW-Y  UP-Y
Miami UP-Y UP-Y  UP-Y
Western Michigan  DW-Y UP-N  UP-Y
Colorado College UP-N UP-N  UP-N

In general trends, the model was 18-8.

One more – let’s look at predicted rank vs. actual rank in each metric:

Team Pred. Sh%  Act. Sh% Pred. Sv% Act. Sv% Pred. Poss% Act. Poss%
North Dakota 1 4  1 1 3t 6
Nebraska Omaha 2 1 4t 2  3t 7
Minnesota Duluth 5 5 8 3  1 2
St. Cloud State 3 7 6 5 5 3
Denver 4 2 2 6 8 5
Miami 8 3 3 4 2 1
Western Michigan 6 6 4t 7  7 4
Colorado College 7 8 7 8 6 8

Kind of a mixed bag here, suggesting a higher level of variance than expected in year-to-year adjustments.

So while the actual percentages weren’t terribly accurate, we could predict trends and relative quality of teams based on previous performances. I think there’s an important distinction to make: it appears we can predict final standings based on shot-based metrics, but we need a better method of predicting shot-based metrics preseason. That is, the model is fine but the inputs need to get better. “Trending up” or “trending down” isn’t accurate enough. Last year’s data and roster imputations will only go so far. That’s something I’ll try to remember for next year. And, of course, as we get a few more years of NCHC games, these things will get more accurate, more reliable, and more significant.

Until then, the experiment continues. This will probably be my last model write-up until the end of the season, unless something major comes up. Meanwhile, I’ll update the model weekly at joelgehringer.com/thelab/NCHCmodel, so bookmark it and check in weekly to see where shot-based stats say your team will end up.

Leave a Reply

Your email address will not be published. Required fields are marked *