Thus far in these weekly Staturday posts, I’ve focused on player-level stats. A new angle I’d like to take today is to look at team-level stats, to see just how good the Leafs have been this season.

I want to frame this post with two key questions that I aim to answer:

Unlike quality player-level data, where Natural Stat Trick (NST) and Evolving-Hockey (E-H) are our the main resources available, there’s a bounty of websites that host team-level Corsi data of value. Adding to that tandem of stats sites, I’ll also be using Puck On Net (PON), Money Puck (MP) and HockeyReference (H-R) for today’s post.

If you’re new to these kinds of statistics, check out my primer.

This season, basic Corsi shot attempts look fairly bad for Toronto. From NST, at 5-on-5 and Even Strength, they’re 25th and 27th in the league respectively. That’s a quite ugly place to start, but also a very simple picture of what’s happening

One way we can add some detail to this is to apply adjustments to the stats. First, there are Score Adjustments, which pretty much everyone can agree make sense. When leading heavily, teams tend to allow significantly more shot attempts, as they begin to believe the game is in the bag, or drop back into a defensive gameplan that bleeds low-quality attempts.

Second, there are Venue adjustments. Each team has an official stat keeper, who has biases. When you play half of your games at home, those biases can show up in the data and unfairly credit a team’s performance at home.

Third, there are Event adjustments. These are less common but I still find them interesting, because they’re a bit counter intuitive. Basically, the Event adjustments gives more value to events that happen less often, which is meant to give more credit to shots that are goals than shots that are not goals. However, more shots hit the net than those that miss or are blocked, about double in fact from PON’s explanation, so in Event adjustments shots on goal are actually valued less than misses or blocks. The above-linked article shows that even though this is counter-intuitive, this adjustment does provide some increase in predictivity (with a decrease in repeatability).

At PON, these stats are applied in layers on top of each other, as follows, shown for Toronto only for simplicity. All of these are at 5-on-5.

The main reason they’re better when looking at score adjustments than basic Corsi is that they are near the top of the league (7th) at Corsi when the score is tied, which Score Adjustments value the most.

It’s also interesting that Venue adjustments creates no difference, meaning that Toronto’s stat keepers are fairly balanced in tracking their shot attempts, so Toronto’s shot attempt stats aren’t unfairly biased towards them.

Lastly, the Event-Score-and-Venue adjustment positively affects the Leafs even more than the Score adjustment alone because, as we know, Toronto scores a lot of goals.

We’ve discussed what Expected Goals (xGF%) are a number of times before, but to quickly reiterate: Expected Goals are a statistical model that value shots based on where they’re taken from. The closer to the opposing net your shot for is, the better. Similarly, the further away from your net your opponent’s shot is, the better.

We’ve also talked about how Corsi is more valuable recently than Expected Goals are (2020), so keep that in mind.

Expected Goals can be a nebulous thing because once you start adjusting the data, you’re giving way to the biases of the person(s) applying the adjustment. As such, I’ve included three different Expected Goals models and how they compare with the CF% from that site, to see if there are differences. All of these are at 5-on-5 with no score, venue or event adjustments.

As we can see, the xGF% largely favours the Leafs relative to basic Corsi, at different amounts depending on the site you use.

Goals Above Replacement, or GAR, is another statistic we’ve talked about a lot, specifically the GAR model from E-H. I won’t discuss it in detail here, but instead point you to part 2 of my advanced stats primer where I explain it in detail. Succinctly: GAR is a model that shows how a team is performing, using a bunch of different statistics rolled into one model. There is a second model, Expected Goals Above Replacement (xGAR), that predicts future GAR. The way this works is that GAR is meant to show the results of your performance, while xGAR is meant to show what results you should expect in the future.

My hypothesis here was that the Leafs would have excellent GAR and terrible xGAR, but I was pleasantly surprised. The Leafs are 1st in the league in GAR, as expected, but are surprisingly 2nd in the league in xGAR.

Components wise, the Leafs are mostly gaining their success on the powerplay where they lead the league by a healthy margin in the xGAR model. The remaining components, even strength offense & defense, shorthanded play, takeaways, faceoffs and goaltending are all in the top half of the league as well, which leads to this 2nd place standing.

Well, it’s incredibly rare for the team at the top of the standings to also be leading in shot attempts. Shot attempts are not meant to say whether you should or shouldn’t have won particular games. They’re meant to show what may happen in the future, which can be interpreted as the quality of the team.

As such, with such mediocre Corsi numbers, even though we can dress those numbers up in different make-ups to make them look better, are concerning. We could be in for a drop in results.

The Leafs are excelling at scoring goals, as we know well. Specifically, they are generating scoring chances, and are dominating the shot attempts when the score is tied.

They’re 5th in the league in scoring chances for per NST, and as we showed before they are also 5th in the league in Corsi % when the score is tied per PON.

The GAR model at Evolving-Hockey absolutely loves our Maple Leafs, and while its reliability is questionable, it’s still valuable that they sit so highly there.

Also, individually, each of their goaltenders has been great. Frederik Andersen is 6th in the league in Goals Saved Above Expected among goalies that have played at least 10 games with 3.6 GSAE. Campbell and Hutchinson are also both positive with 5.0 and 0.6 GSAE respectively in their 3 and 5 games. This a stat that at a team or skater level generally means luck, but for a goaltender it means good performance. There’s a nuance there to balancing those two perspectives for the same number that’s beyond my capability, but I just want you to be aware of it.

The Leafs lead the league, by a healthy margin, in our “luck” statistic, PDO. They’re getting a shooting percentage of 10%, which isn’t really unusual for the Leafs, but the goaltending has been unsustainably good at 93.66% at a team level. Time has shown us over an over again that numbers this high regress towards more normal numbers, which for team save percentage is more like 91%.

Their defense appears to be the best it’s ever been, but PDO clearly shows us that this is being propped up by unusually good goaltending. When, not if, this prop falls, then there will be standings consequences. Whether this will happen this season is impossible to know. All we know is that the future looks not-so-good for Toronto, by these numbers.

Finally, the Leafs are not generating enough shot attempts for. They sit 6th worst in the league in Corsi For per 60 minutes of time at 5-on-5. While the larger percentage of their shot attempts are scoring chances, the data shows that even if that’s true, total shot attempts of all kinds are important to control.

The sky certainly isn’t falling, but there is rain in the forecast for the Maple Leafs. I want to avoid making sweeping proclamations that the Leafs are in for a rough go down the rest of this season. We’ve seen many times that PDO can stay afloat for a whole season and come crashing down in the playoffs.

All that this data should present is a red flag. It’s not a sign of imminent danger, but something to be cautious of when evaluating the team’s future and their chances to make a long run in the league.

It’s common for those who are involved in advanced stats to be wholly pessimistic, using the data only to show where others are wrong, and especially when teams and players who are beloved might actually be quite poor. I’ve been mostly positive with these columns, and I don’t aim to bring this contrarian way of writing to these columns. However, negative numbers are still important to show just as much as positive ones. I hope that my reasoning for taking this negative perspective today is understood: because that’s the way the numbers are presenting themselves to me.