For those following the discussion of the Islanders this past season online, particularly the discussion in analytics circles, there’s been a serious question of whether the Isles have been playing over their heads. In fact, most models and analysts did conclude that the Isles were getting fluky results to some extent last season, and that a cool down of some sort should be expected, perhaps from their goalies coming down to earth or from their overall shooting numbers declining.
To make it clear here, I’m not talking about typical analyst snobbery about the Islanders as this small market team who no one watches. I’m talking about how the Islanders as a team were decently outshot (Corsi) this past year, even when adjusting the numbers to reflect that the Isles spent more time leading than most.
My colleague on this site, Carey Haber, has posted more than a few times on a discrepancy in the more “advanced” stats that might show the Isles were better than those others suggest. In terms of Expected Goals, which weight each shot based upon location and likelihood of being a rebound shot (and occasionally other factors like type of shot), the Islanders were actually an above average team, outperforming opponents by a little bit. Carey has suggested and argued that this was the result of Barry Trotz’ system,
Similarly, there has been a debate over whether or not the performance of Islander goalies - last year by Robin Lehner and Thomas Greiss, or this next year by Greiss and Semyon Varlamov - was the result of unsustainable performances, or something to do with Trotz and longtime confidant and goalie-whisperer Mitch Korn. With the Isles bringing back substantially the same team and not bringing back their most prominent goalie, these questions loom large as we look at whether the Isles are again a playoff contender.
Note: All Data is from Natural Stat Trick.
The Trotz Effect on Expected Goals:
What we’re really talking about when we discuss “Corsi” vs “Expected Goals” is shot quality vs shot quantity.
It has long been found by analytics work in hockey that, when adjusting for the score at least, one of the best predictors of future team success is whether or not a team is putting up more shots than their opponents. That’s all “Corsi” is - a measure of what percentage of shot attempts are being made by the team you’re looking at. It’s not very complex, but it works wonders, and shot quantity has allowed us to predict which teams are for real and which are pretenders destined to drop back (and which are potential sleeping giants) for years now.
Expected Goals are a newer addition to the Hockey Analytics Lexicon, but while they’re complex in calculation, they’re pretty simple to understand in theory. Expected Goals are simply an attempt to answer the question of “Shot quality,” adjusting our simple counts of shots for and against by how dangerous each of those shots are. A team that’s getting more dangerous shots on average than their opponents may be able to overcome getting outshot, which was what happened with the Islanders in 2018-2019.
But in general, the gap between Corsi and expected goals for teams over longer periods of time tends to be small. That’s not the case with the Isles for this past year. Their score-adjusted expected goal percentage (50.44%) was substantially higher than their Corsi percentage (48.47%). And if you’re familiar with sports analytics, you may know that weird outliers that don’t fit the usual tend to raise red flags about being unsustainable, and are unlikely to continue in the future.
But what about Barry Trotz, Jack Adams Winner? Fortunately, its not like Trotz is a coach with a small track record. He’s been in the league seemingly forever at this point without any breaks, over now three different teams. So, I decided to look at Trotz’s past 10 seasons (one with the Isles, All four with Caps, and the final five in Nashville) to see if there were any trends in this direction we could extrapolate.
So what are we looking for in this study?
- Whether Trotz-coached teams can consistently win the shot quality battle enough to make a substantial improvement on what would be expected from their shot quantity numbers.
- If Trotz has had outlier seasons like the 2018-19 Isles, and what has been the result the year after?
The answer to the first issue is a qualified “yes,” although not a big enough “yes” to make Isles fans feel particularly hopeful.
Over the past 10 years, Trotz’s teams have managed a score adjusted Corsi of 49.78% and an expected goals percentage of 50.57%. That’s an improvement of around 0.8 percentage points, far lower than the 2.0 percentage points the Isles managed in 2019. If the Isles’ shot quality over-performance drops accordingly, the Isles are going to see a significant drop in overall numbers, or a couple wins worth, which in the tight Metro Division may be too much.
But that analysis oversimplifies things a little bit. In Trotz’ last five years with Nashville, he managed an average over-performance on shot quality of around 1.5 percentage points (50.3% xG%, 48.8% CF%). But in his more recent tenure as the Caps head coach, the trend went the other direction, with the Caps LOSING the shot quality battle, allowing more dangerous shots than they managed on the offensive end, resulting in an Expected Goal Percentage lower than their Corsi Percentage. (50.9% xG vs 51.2% corsi).
This may seem confusing, since no one thinks of the Caps as taking worse shots than their opponents, no matter who they are, especially with their personnel. This is where I remind you that Expected Goals takes into account location, type, and timing of shots, and not who is doing the shooting (and the Caps have one of the deadliest shooters known to all of hockey). Also, the Expected Goals numbers we have do not account for pre-shot movement that make shots more dangerous - one-timers and the like - and recent research shows that the Caps may indeed excel in this area. Finally, the Caps did a lot of their damage on the power play, which I think we all can admit is not something the 2018-19 Isles were good at. Similarly, the Isles do not have Alex Ovechkin to pull off the same shooting feats as those Caps, and while the Isles have some good passers, that’s not the over-performance we’re talking about here. So whether that aspect of Isles performance changes or not is beyond this article’s scope.
But the point is that if we assume that somehow the Isles’ over-performance in shot quality was due to Barry Trotz’s system, it’s odd to find that that over-performance was not seen in Washington. And even averaging in Nashville’s numbers suggests that we can’t expect over-performance to such a degree to continue on the Island.
Which is not to say that this is the first time that a Trotz team has over-performed so. As I noted above, it happened to Trotz’s last three Nashville teams to an even greater extent than the Isles! The thing is however, that not only was that a long time ago now - five years past - but those last three Nashville teams weren’t good, with two of them being worse than this year’s Isles and getting Trotz replaced in the process. They over-performed from being dreadful (47% Corsi) to merely average (50.2% expected goals), which is not exactly a reassuring place to be for a team with lousy special teams and a tough division.
The Trotz/Korn Effect on Goaltending
Okay, so what about Trotz and Mitch Korn’s effect on his goalies? Well, believe it or not, the data sample last year’s performance was not the first time a Trotz team has put up an insane save percentage. In fact, it’s not even the best one for a Trotz team: last year’s 5-on-5 SV% of .9366 was bettered by the 2016-17 Caps (.9373) and almost equaled by the 2010-11 Predators (.9345).
You will note however that those years are pretty far apart and, in the seasons before and after both of those other insane goaltending performances, Trotz’ teams put up much more normal save percentages, if still good: In 2011-12, the Preds, for example, dropped from .9345 to .9257, while in 2017-18, the Caps dropped from .9373 to .9248.
This shouldn’t really surprise anyone. It’s just basic regression at work as players perform over their heads and fell down to Earth. There’s no noticeable effect of Trotz being able to maintain superior goalie save percentages, even without star goalies, in the ten year sample I’m using, which wouldn’t be a positive sign for the Islanders even if they had retained Lehner. And Varlamov is almost certainly a worse goalie at this point in his career.
Obviously, nothing in the above takes into account changes in the Isles’ talent levels from year to year, so nothing in the above should suggest that “OMG, All hope is lost!” Mathew Barzal could easily take another leap, younger Islander players finally given playing time (Michael Dal Colle, Josh Ho-Sang, Otto Koivula, etc) could improve the underlying numbers, etc.
But for the most part, the Isles are bringing back the same team as last year, and a major reason for last year’s team’s success was the over-performance described above. And Barry Trotz’ past results don’t suggest that such over-performances are the norm, which suggests at least somewhat of a decline that needs to be made up for elsewhere.
So don’t be too surprised when you see projections showing the Islanders taking a step back....because a look back suggests they may be very unfortunately right.
For those curious to see the data involved in this post, you can find it linked HERE. As noted above, all data is from Natural Stat Trick.