The first round was very difficult to predict. Are there more upsets in store for the second round or does our prediction model see how it will play out?
If this year’s first round felt more unpredictable than usual, it’s because it was. When the playoffs start, it’s a brand new season. Nothing else matters and anyone can win a series. That’s one big lesson from Round 1.
Chicago, the top team in the West and an extremely popular choice to make it to the final is out. Minnesota, the second best team in the West and a popular choice from the analytics crowd, is out. Montreal, a team that finally looked like more than its superstar goalie, is out. Boston, the league’s best possession team, is out. San Jose, last year’s West champ, is out.
It’s been a wild and chaotic start to this year’s playoffs with four major upsets already in the books (not counting the Sharks). The numbers crowd “called” one of those: Nashville beating a much more vulnerable Chicago team wasn’t all that surprising, but doing it in four games sure was. No one called that.
But while the nerds deserve credit there, they also underestimated Ottawa’s chances against Boston relative to mainstream media, as well as St. Louis’s against Minnesota. In both cases, there’s a reasonable excuse to be made (Boston was decimated by injury, Minnesota was decimated by Jake Allen’s best brick wall impression), but it doesn’t change what happened on the ice. Both camps were equally high on the Habs, so that’s a wash.
The model we used here was a bit kinder to Ottawa, but was even higher on Minnesota and Montreal. Overall, it didn’t perform very well in the first round, but that’s what happens with so many upsets. Based on log loss, a method of measuring prediction accuracy, this year has been much less predictable for everyone.
Last year, based on all the predictions I collected, the average log loss was 0.619. This year it’s 0.711, which is really weak considering calling every series 50/50 would get you to 0.693. Last year, every prediction method but one was above that. This year it’s only 15 of 35, with most of those coming from the pundits, but even they’re worse on average. It’s been a very humbling experience for a lot of expert prognosticators, but especially for the stats crowd, who are regressing hard after an excellent 2016 playoffs.
What happened on the market should say it all about this first round. Pinnacle, the sharpest bookmaker you’ll find, was more accurate on their opening lines than they were on their closing lines. Lines generally move toward big money from sharp bettors and they’re sharp for a reason: they win. It’s extremely rare for the openers to be more accurate than the closers over a large sample, but that’s what happened so far. Given that, and the fact there’s only been eight series, I wouldn’t put much stock into evaluating anyone’s picks just yet.
Predicting the playoffs ain’t easy. That’s the other lesson from Round 1.
Nevertheless, we’re going to keep doing it anyway. Upsets happen. It doesn’t mean you should throw out everything you believe in, which in this case is a model that performed very well during the regular season. It’s why we deal in probabilities, not absolutes. A 60 percent chance doesn’t mean 100 percent – it means the other team still has a very real chance, just not as good as the team on the other side.
Here’s how the model believes the second round and the rest of the playoffs shake out as well as some thoughts on why.
1. There’s about a 50 percent chance that the Stanley Cup goes to either Washington or Pittsburgh. That sounds about right to me. No team was really on their level going into the playoffs and that remains true now, especially with some of the next contenders in line bowing out in the first round (and relatively weak opponents waiting for them on the other side of the bracket).
This is the real Stanley Cup final right here and I have little doubt that these are the two best teams in hockey. The fact they have to meet each other in Round 2 instead of three while the Sens and Rangers play each other is a travesty.
2. Washington is still the clear cut favorite to win it all and are, perhaps surprisingly, given a much better chance to move on over other teams, despite the tough match-up. If Washington is at 59 percent over Pittsburgh, you can imagine what they’d be at over anyone else.
The consensus from most models seems to be in that 60/40 range (for the record, it was mostly the opposite last season in favor of Pittsburgh), but it seems like the majority of people see this as closer to a coin flip. I wonder how much recency bias plays a factor in that, because the Capitals barely survived against a scrappy, young, Leafs team while the Penguins obliterated Columbus who had the fourth most points in the league this season. Things change from round to round and Washington should be better prepared to face their rivals this time around.
The key here will be on the backend. Matt Murray would have made things much closer, but Braden Holtby vs. Marc-Andre Fleury is a bit of a mismatch, especially given the defense supporting both guys. Washington is extremely deep back there while the Penguins look like they could get exposed. I wouldn’t be surprised if Pittsburgh wins, again, but this really does feel like Washington’s time.
3. Perhaps the most surprising thing here is Ottawa being favored in their match-up against the Rangers. Ottawa. Favored. To go to Round 3. According to the guy who repeatedly called them a first round guppy all year. Whoops?
I meant every word I said all season. Ottawa was a bad team getting some breaks. Was. This is a very different group than the one that won games inexplicably in the first half of the season. Their bottom six was atrocious, but adding Alex Burrows, Viktor Stalberg, and Tommy Wingels has pushed some very bad players out of the lineup. Then there’s the Clarke MacArthur factor. That’s a big time addition right before the playoffs that very few models can account for. He’s a good player that’s already scored some big time goals these playoffs.
All those depth additions add up and they’ve suddenly got four actually capable lines. They’re led by Mark Stone and Mike Hoffman up front, two very talented and underrated wingers who I believe could arguably be the best forwards on the Rangers. Combine that with the fact they have the league’s best defenseman on the backend and Craig Anderson playing his best hockey in years – suddenly it doesn’t seem all that crazy for Ottawa to win. This current iteration is a slightly above average hockey club, and that could be enough to take down New York.
4. Okay, let’s be honest. All of that is a little crazy, right? The Sens favored over the Rangers doesn’t make much sense. The few mainstream media predictions I’ve seen have been overwhelmingly in favor of New York and I don’t blame them. The models are a completely mixed bag though and there’s zero consensus right now other than the fact most agree it should at least be very close. Of the 19 models I’ve seen, 10 have Sens, nine have Rangers. At least I’m not alone.
With these two teams, logic and reason don’t seem to matter though. This is the Anti-Analytics Bowl, a match between two teams that have defied the Corsi Gods all season. Why would any prediction between these two teams matter now?
5. Still, this is probably the least confident I’ve ever been in a prediction. I think the Rangers are a lot better than my model currently gives the credit for, and are likely better than their regular season numbers indicate (though still not as good as most people think they are). They’re the worst possession team here, but even by goals they weren’t that good at 51.6 percent, 12th among playoff teams. There was good reason to be skeptical besides Corsi because they weren’t out-scoring teams by much at 5-on-5 either.
The issue there was Henrik Lundqvist. Their shooting percentage was still as high as usual thanks to their counter-attacking style off the rush (which is why I think current data might underrate them), but Lundqvist’s play wasn’t as pristine as usual. He only managed a .910 save percentage this year, the lowest mark of his 12 year career. He’s really getting up there in age, but I doubt he’s that bad yet. With him at his usual .920, the Rangers look much better.
That career low drags him down in my model. When you expect a .916 (using his last three seasons) and he puts up a .947 instead, there’s a good chance the original prediction won’t work out very well. Models can’t account for Lundqvist reverting back into The King. And if The King is back, I won’t be betting against him.
6. So how does “Nashville Predators, Official Dark Horse Team” feel now? They were exceptional against a Chicago team many thought could contend for another Cup, giving them sweet nothing all series. Pekka Rinne isn’t maintaining that .976 (!) save percentage this series I can tell you that much, but even without his super-human effort, the Preds were in a very good spot. They dominated the Hawks with 56.3 percent of the expected goals that series, second best for the playoffs so far. To do that to that team is impressive. So while Rinne was obviously a wall, Nashville earned their win.
With Chicago and Minnesota out, I think they’re the team to beat out West. They’re probably not beating either of Washington or Pittsburgh, but they’re next in line and Anaheim is close right after. There’s a pretty substantial drop off after that top four in this model’s opinion.
The key for Nashville is that no one can stack up with that backend, their top four is just unfair. Their top line is working some magic right now too, that’s a very underrated trio they have in Ryan Johansen, Filip Forsberg and Viktor Arvidsson. They also get a relatively easy match-up in the second round. By my model, this series against the Blues looks to be the most lopsided in terms of talent.
7. The Blues won the first round battle, but they’re not winning any wars playing the way they did. I was a little nervous about calling them one of the weaker teams going into the playoffs, but even after seeing them win four of five, I feel pretty confident in saying that analysis was spot on. Allen was magnificent, but the team in front of him was a disaster. Earning just barely more than 40 percent of the shots in a series and relying entirely on goaltending is not generally a sound path to success. Stranger things have happened and Allen could keep his stellar play going, it’s just not something I would bank on.
I do think they’ll get more chances this time around (they can’t be that bad again), but Nashville should still be able to carry play here while both goalies likely turn back into pumpkins. Some modest regression from the two incredible goalie performances should make for a really fun series.
8. One last series to touch on and that’s in the Pacific. Anaheim seems to be a consensus favorite here and it’s not hard to see why: more depth everywhere. Edmonton has the Connor McDavid factor, but outside of him this looks like a match-up nightmare for the Oilers. The Anaheim top nine is stacked with talent and their defense is deep. I don’t mind the Oilers support guys, but I don’t think they’re at Anaheim’s level. I just can’t picture Ryan Nugent-Hopkins, David Desharnais or Mark Letestu matching up favorably against either of Ryan Getzlaf or Ryan Kesler.
What could work for Edmonton is, well, that McDavid factor. He was awfully quiet, for his standards, during the first round with just four points in six games and if I were Anaheim I’d be just a little nervous of a market correction for him. Very few players can carry a team on their own, but McDavid is definitely one of them. If he gets hot, watch out for Edmonton.