Keefe’s Leafs: what is this team doing differently?

Ryan Hobart
1 year ago
After two incomplete seasons as the Leafs’ head coach, it feels like Sheldon Keefe has been around here a long time. But at the head coaching position in the NHL, he came into this season with just over 100 games under his belt, a little more than 1 full season.
Now, 2 games into the year, we’re already hearing rumblings of whether he’s the right fit for this team. In those partial seasons, Keefe’s Leafs had a better in-season record than any of Mike Babcock’s Leafs. Of course, there’s an asterisk on the 2020-21 season because they got to beat up on the Ottawa Senators for a higher percentage of their games than normal (notwithstanding Thursday night’s disappointment). Still, his overall points percentage with Toronto is 65.7%, a fair notch higher than Babcock’s tenure here with 55.7%.
What I’m curious about, which perhaps you’re curious about it too which and that’s why you’re here, is: what is Keefe having this team do differently at 5v5, and how does that bear itself out in the numbers?
The head coach is the primary influence at 5v5, while assistant coaches generally run the special teams on the powerplay and penalty kill, so that’s why we’re going to focus on 5v5. Conveniently, this is also when the data is the strongest.
I’m going to compare Keefe’s 2020-21 season to Babcock’s 2018-19 season. We can see pretty obvious changes in the 2019-20 season when Keefe took over, but that’s less relevant now than the 2020-21 season.
Looking at Alison Lukan’s coverage on this exact same topic before the 2020 NHL playoffs, there were a two conclusions drawn about the offense:
  • Increased ice-time for @Auston Matthews, @John Tavares, @Mitch Marner and @William Nylander
  • Permanent marker on the team sheet for the 1st and 2nd lines, for the most part
Otherwise, the article is paywalled, so we will draw our own conclusions with what’s publicly available, and also test if those trends are still true.

Forward Deployment

First, let’s look at the ice time share for last season:
We can see that at the start of the 2020 season, Keefe started with his ice time share being similar to how it worked in 2019-20: lots of ice time for Tavares, Matthews, Nylander and Marner. However, it very clearly trended downward as the season went on. Part of the reason for this is probably that Matthews was suffering from a wrist injury, which he’s now recovering from the surgery for. This coincided with a big jump in @Alexander Kerfoot’s ice time to compensate.
What about the combinations? Alison concluded that Keefe had intentionally stapled the pairs of Matthews/Marner and Tavares/Nylander, more so than Babcock was at the start of the 2019-20 season. Here’s how the network of forward ice time looks when compared with Babcock’s 2018-19 season versus Keefe’s 2020 season:
In 2018-19, the Hyman-Tavares-Marner line was a staple that almost never separated, while Matthews had a variety show of linemates. In 2020, Keefe had Matthews and Marner together almost inseparably, and similarly Tavares and Nylander were almost just as glued together. However, their third linemate for each of those lines varied quite a lot. This is fairly similar to Alison Lukan’s conclusion above from the 2019-20 season.
What else can we learn? Here’s a chart showing forward deployment depending on what the score of the game was. Moving from left to right, on the left we can see how forwards are used when Toronto is losing and on the left we can see how they’re used when Toronto is winning:
I would ignore the “-3 or more” side on the far left, because we can see that only ~1% of Toronto’s ice time was spent at that state, so the variances here are massive because of small sample size.
In the other sections of the chart, we can see some sensible trends for Babcock’s team in 2018-19. The top forwards get more icetime when they’re losing, and less ice time when they’re winning, but overall, the ice time distribution is pretty consistent.
Looking at Keefe’s chart for 2020, the deployment for Matthews and Marner literally goes off the chart when Toronto is losing by 2 goals. The other major difference is we can see that Nylander’s ice time went way down when the team was winning, and Hyman’s went way up. This leads to conclude that Keefe makes the score of the game a much larger factor in determining who will be on the ice than Babcock did.
It’s far too early to look at how this deployment is shaping up in 2021-22, especially with Matthews out of the lineup, but we’ll take a look at that later on in the season and see if these trends hold true.
To sum up, here’s what we discovered:
  • Keefe has solidified the pairs of Matthews/Marner and Tavares/Nylander consistently, while Babcock used Hyman-Tavares-Marner consistently while the rest of the roster varied
  • Ice time distribution used to heavily favour the top 4 players, but trailed back to a more even distribution as Keefe has spent more time with the team
  • Keefe’s deployment changes more dramatically based on the score of the game than Babcock’s

Defense Deployment

The defense usage is pretty similar when comparing 2018-19 and 2020-21, so it’s not very interesting, but it’s still important to know that it hasn’t changed. Here is the ice time networks, showing three consistent pairing combinations:
One thing that’s interesting is the usage of @Morgan Rielly and his defense partner. In 2018-19 he was paired with @Ron Hainsey, but when the team needed a goal, the ice time Hainsey and @Nikita Zaitsev went down, and Rielly and @Jake Gardiner were the ones trying to drive offense together from the back end. We also saw Travis Dermott being used more when the team needed a goal.
Now looking at 2020-21 with Sheldon Keefe, he’s pretty consistent with the ice time distribution regardless of the score, which is opposite to how he uses his forwards. Rielly gets a bit more ice time when the team is losing, @Justin Holl gets a bit less, but overall it doesn’t change that much.
Unfortunately, there’s not much else to conclude about the defense in terms of usage.

On-Ice Performance

Other than their records, which Keefe has the better as shown in the intro, we can also look at how the team performed on the ice independent of results by using shot attempts. This is one of the core concepts of analytics in hockey: more shot attempts for and less shot attempts against is a recipe of future success. Adding another layer to that, we can look at where those shot attempts come from, and give the ones with more chance to score a higher value. This is called “Expected Goals” and is what will be shown in the heat maps below.
Let’s look at offensive performance first. Here’s Babcock’s shooting performance from the 2018-19 season:
And here’s Sheldon Keefe’s from 2020:
The only real difference is that Toronto’s offense is more centered at the net front under Keefe, and more from the circles with Babcock. These are both dangerous areas, so neither is really better or worse, just strategically a bit different. The overall performance in Expected Goals For per 60 minutes of ice time (xGF/60) is pretty much the same.
Now on the defensive side:
We can see that in Keefe’s 2020 season, the Leafs defended the front of their net really well. Both teams were susceptive to shots from the inner circles, but much more so in Babcock’s 2018-19 team. In overall performance, Keefe’s team is decidedly better, having a negative impact on the Expected Goals Against per 60 minutes of ice time (xGA/60), meaning they gave up fewer chances than league average. Conversely, Babcock’s 2018-19 team gave up more than average. However, -6% for Keefe isn’t a glowing result, generally for a top tier team you want to be a bit stronger than that.

Final Thoughts

We know that the team is doing better under Keefe. That much is clear. But it’s also obvious. Babcock’s Leafs were flattening as it came to the point where he was fired; that’s why they fired him. To improve on that is the baseline expectation. It doesn’t seem obvious that Keefe has the team doing anything exceptional.
There are some differences between Keefe and Babcock, as could be expected, but I don’t think there’s much evidence that Keefe is taking this team to another level. Could he? Is there some limit of diminishing returns that the Leafs have hit in terms of regular season performance? Maybe, but looking at the Lightning or the Avalanche, I would say no.
Until Keefe finds a way to elevate the non-Matthews big four forwards to dominant play, this probably isn’t a Cup-winning team. But with how Nylander has played lately, and how insane Matthews has been for the last 2 years, maybe this is finally the year they get past the first round.

Check out these posts...