Photo courtesy of the Ithaca Journal.
Over the weekend, teams battled it out across the Northeast in the the ECAC quarterfinals. We saw some great matchups, including top-ranked Harvard knocking off arch-rival Yale, second-ranked Union going into overtime to beat Princeton, and Quinnipiac rallying to upset St. Lawrence.
The big story for us, though, was obviously the Cornell-Clarkson series which may have been the best of them all. After being thumped 6-2 in Game 1, Cornell won Games 2 and 3 by 2-1 scores to clinch their berth in the semifinals next week against Union.
Why was the series, which was between a top-10 team in the country and an unranked opponent, so close? How did Cornell play so poorly as to lose Game 1 by that much, and why were the other two games so close as well? Cornell is expected to receive an invitation to the NCAA Tournament, so the fact that Clarkson came so close to knocking them off may be surprising to some.
There is, however, a fairly straightforward explanation for how close the series was. You may remember an article published last week detailing advanced statistics and their use in both the NHL and the NCAA. Specifically, I focused on a number known as Corsi, or a team’s percentage share of shot attempts in a game. As I mentioned, Corsi has been shown to be a very powerful predictor of future goals and, as such, can often be used to measure a team’s success. I showed how some of Cornell’s trends were worrisome, specifically their Corsi score below 50% (meaning they tended to be the worse team in the games they played) and were getting quite lucky, and that we could expect their play to regress a bit.
I think we saw some of that in Game 1 when Clarkson exploded with 6 goals. Considering that Cornell’s goals-against average in the regular season was just over two, this was an aberration for sure. On the other hand, it did show some regression. Cornell’s PDO, or combined shooting and save percentage, in that game, was 93.981, far lower than either the 100 PDO mean or Cornell’s regular season PDO of over 102. While that number is certainly out of the ordinary, and a product of the game in question, the 5v5 Corsi was right on par: 38.57%. Cornell was completely out-played at even strength, which means it is not very surprising that they fell by that large of a margin.
Fortunately for Cornell, the regression did not last the rest of the way. Goaltender Mitch Gillam regained his form, Clarkson goalie Jake Kylie lost his, and Cornell’s PDO skyrocketed. The combined PDO for Games 2 and 3 was an absolutely astronomical 109.825! There is absolutely no way that value will be sustainable; heck, even if you believe Cornell’s already high PDO of 102 can last over the long haul, a value of nearly 110 cannot last more than a couple of games. It is worth mentioning that Cornell was badly out-attempted at even strength in both of those games, although I cannot provide a Corsi value since I do not have the raw numbers with me.
All of this is to say that, while Cornell may have gotten the luck it needed to get by Clarkson, they will need to perform significantly better than we have seen this season to hope for a chance going forward.
Elsewhere in the ECAC
As I mentioned, the other three matchups saw Harvard defeat Yale, Union beat Princeton, and Quinnipiac upset St. Lawrence. Having read the details about the Cornell-Clarkson series, you may be wondering if the other results match up to what the statistics predict. After all, maybe Cornell was not alone in bucking the analytics, in which case all of the points I made are moot.
Well, it just so happens that Cornell-Clarkson was pretty much the only series that saw team with poor analytics win its series. Let’s start with Harvard-Yale. Harvard has a 5v5 Corsi% of 53.5, which is very, very good. Yale, however, sat even better, at 54.4. While this technically violates the prediction made by Corsi, if you look at the actual games between the two, a different story emerges. Harvard outshot and out-attempted Yale in both games, and the fact that the two games were so close (6-4 and 4-3) seems to prove that these two teams really were evenly matched, and the difference in this series simply came down to a few extra shots one way or the other.
The next series is Union over Princeton. Union, like Cornell, is a somewhat poor Corsi team, especially compared to Harvard and Yale, with a number of 50.6%. Its opponent, Princeton, had a Corsi of exactly 50%. As such, we could have expected this series to be a toss-up, and it sure was. Game 1 was won by Union 4-1, but one of those goals was an empty-netter, and another was on the power-play. At even-strength, the score was 2-1 Union, which correlates well with the numbers. Game 2 really was a toss-up, with Union prevailing 4-3 in overtime. Even if you take out the two power-play goals by Princeton and the 6v5 game-tying goal and penalty shot game-winner by Union, the even-strength score was again 2-1. This once again correlates with what we would have expected from Corsi, a close, tight series.
The final series was Quinnipiac over St. Lawrence, the lone upset of the quarterfinals, and this is the series where Corsi is shown in its fullest. Quinnipiac is one of the best teams in the league with a 55.9 5v5 Corsi, whereas St. Lawrence is on the opposite end of the spectrum, with just a 47.0 Corsi number. Although the series was close (all three games were decided by one-two goals), it is no surprise that Quinnipiac could play three games on the road and still win the series.
Overall, Corsi aligned well with the results in the ECAC Quarterfinals, Cornell-Clarkson excluded.
Lake Placid Preview
This weekend, the semifinals and finals of the ECAC tournament take place in Lake Placid. There is no three-game series this time, so Cornell cannot afford another performance like they had in Game 1. Union is probably their best potential matchup, since they are not an elite possession team either (Cornell is currently at 48.4%; Union is at 50.6%). We can expect this game to be close, and it is really difficult to predict a winner, considering how volatile Gillam has been at times this season. Gun to my head, however, I would predict Union to win the game 2-1 or 3-2.
The other game is between Harvard and Quinnipiac, and this should be a dandy. Both are among the league’s foremost possession teams, with Harvard owning a Corsi% of 53.5 and Quinnipiac rocking a sterling 55.9. I would expect this to be a high-scoring and evenly-matched affair, but in the end I think Harvard’s talent makes up for its deficit in Corsi, and they win 5-4 or 6-5.
In terms of the final, I feel comfortable predicting that, regardless of who exactly the winner in each game is, the Harvard-Quinnipiac winner will defeat the Union-Cornell winner decisively in the final on Saturday.
Comments