clock menu more-arrow no yes mobile

Filed under:

Checking the eye-test: A journey through the Kadri line’s results

I wanted to check to see if what I saw was real. And then I had to do it again when I was looking at the numbers.

NHL: Toronto Maple Leafs at Winnipeg Jets Terrence Lee-USA TODAY Sports

In Monday’s game against Chicago, in the opening minutes, Nazem Kadri took the puck up the ice with Patrick Marleau on his left wing and Leo Komarov on his right. This was a three on two rush, and it was part of the exciting heavy pressure the Leafs piled on rookie goalie Anton Forsberg early on. Kadri passed the puck to Komarov and Forsberg made an easy save.

I snidely said, “No one ever expects the pass to Komarov.” And while as a surprise tactic, it might work, but I wonder how much studying of Leafs forwards Forsberg has done. He’s now in the Western Conference after some games with Columbus last year, so perhaps none.

All three players on Kadri’s line are left shots, and he brought that up way back in training camp that it was something they might have to be aware of. While it’s true that positions are not held rigorously in the offensive zone, the idea that it’s all chaos all the time and handedness or winger vs centre is irrelevant is grossly overstating the case. A left-shot centre is going to pass more easily to his right, and that’s going to be the right wing a lot of the time.

But who would you rather have taking that shot? Why didn’t Kadri pass it to Marleau?

After watching all three games this season, I had this niggling feeling that Kadri’s line was not executing in the offensive zone very well. I has a sense that they were not getting good chances. So I set out to see if that was true.

Disclaimer: This is the “thing you did” versus “thing you are” reminder. Three games of data is not predictive of anything. This is 100% things they did. Given that everything I’m about to talk about relies on the NHL-collected shot location data, the quality of that data has to be suspect. The smaller the pot of it, the less good data there is to swamp the junk. So the descriptive value of all of this is less than 100%.

I started out looking at some shot plots of individual players. The only ones I could find were all situations, so that wasn’t ideal, but it did show that Kadri has not been getting very close to the net when he shoots, but Komarov has been. Marleau is a mixed bag, and has scored two goals from close in. Ah-ha, I thought.

So I decided to see how that lined up with expected goals data. Expected goals, as calculated by Corsica, is every unblocked shot weighted for location and type. This is a shot quality measure. And that’s what I was wondering about, not how much is the Kadri line shooting the puck so much as how well they are shooting it.

I started with this basic set of numbers from Corsica, using their Line Combos data for un-adjusted five-on-five only. Note, this and everything derived from it, is on-ice results. This is not individual shooting or scoring, it’s everyone (including defencemen) on the ice together when that line is on the ice. I left out Moore’s line because they played very few five-on-five minutes.

This is likely hard to read on a webpage, so here’s the base data published as a google sheet.

Maple Leafs first three games, forward line stats

Player one Player two Player three GP TOI CF% CF/60 CA/60 GF/60 xGF% xGF/60 xGA/60
Player one Player two Player three GP TOI CF% CF/60 CA/60 GF/60 xGF% xGF/60 xGA/60
AUSTON.MATTHEWS WILLIAM.NYLANDER ZACH.HYMAN 3 27.78 60 77.75 51.84 6.48 75.26 7.69 2.53
JAMES.VAN RIEMSDYK MITCH.MARNER TYLER.BOZAK 3 29.38 59.38 77.6 53.1 4.08 55.05 3.45 2.82
LEO.KOMAROV NAZEM.KADRI PATRICK.MARLEAU 3 23.95 46.81 55.11 62.63 2.51 55.09 2.3 1.88
CONNOR.BROWN ERIC.FEHR MATT.MARTIN 2 10.72 57.14 44.78 33.58 5.6 47.67 2.29 2.52

What I wanted to note here is that Kadri’s line has the lowest Corsi For percentage, and while CF% does not equal possession, it implies it, and can be broadly taken for the share of time a line spends in the offensive zone. Their Corsi For per 60 minutes is very low relative to the other two lines in the top nine as well. In terms of actual results, their Goals For per 60 minutes is the lowest of any line.

Goals have a lot of randomness in them, there are reasons why we bother to look at shot data instead of just scoring, and they’re good ones.. Luck clouds the goal picture, so Expected Goals are a more reliable measure of offensive play. While the Kadri line’s Expected Goals For percentage is nicely above even, and the same as Bozak’s, their Expected Goals For per 60 minutes is the same as the fourth line’s. Ah-ha I said again.

I charted out just the expected goals to see how it looked:

I’ll wait while you gaze with love upon the Auston Matthews line. So while the Kadri line is spectacular at defensive play, their offence is tepid. Note also that Matthews’ line and Tyler Bozak’s line are about equal in defence and the fourth line is no better. A lot of eye-tests will tell you it’s Bozak’s line alone that are a problem defensively, but this simply has not been true.

By looking at this chart, it seems like Kadri’s line, even allowing for their role as a checking line a lot of the time, are not generating offence. Hockeyviz’s competition graphics show that Matthews and Kadri have been enjoying broadly similar forward competition so far, and it is tougher than what Bozak sees. The top nine got opposition defencemen in an even rotation, but Eric Fehr’s line got a heavy serving of top pairs. So don’t make the mistake of thinking Kadri’s line spends all their time against top lines, they don’t.

Something was bothering me though. I felt uncomfortable with my reading of this data, just as I had relying on my feelings from watching the games. I’d just confirmed my feelings! But, I kept thinking, what about that disparity in CF%?

What I wanted to know was how the quality of the play in the offensive zone varied by line. But xGF/60 as shown above is made of both opportunity and execution. You have to get in the zone in order to shoot, and then you shoot better or not, depending on a myriad of factors, some of which xG captures.

At Arvind’s suggestion, I took the percentage of Expected Goals For per Fenwick For. The idea here was to create a rating of how likely a given unblocked shot was to go in the net. (Fenwick is used because expected goals is calculated on unblocked shots only.) In this way, I cleaned out the effects in the above chart of how much offensive or defensive zone time a line had.

We pause again to marvel at Hyman, Matthews and Nylander. What that means is that, when they were on the ice, 12% of the unblocked shots were of high enough quality to produce a goal. You can also call that Expected Fenwick Shooting Percentage and Corsica offers that stat directly (xFSH%). League average last year was 5.5%.

Kadri’s line looks much better in terms of quality of shots taken and allowed, and Bozak’s line looks worse. But suddenly the defensive ability of the fourth line is highly in question. I think Chicago caught them on the ice with their second line, though, so maybe that’s a blip from a small sample of minutes.

The quality of the offence of the Kadri line is equal to the Bozak line. I would think both are a too low, and that Bozak’s should be the higher. But I’d have to run this on last year’s data to really know if that guess is accurate.

As Arvind said, this shows why we rely on CF%. Outside the stars on the left, the offensive quality is very flat. The key to success, therefore, is to get more opportunities to score. Although I would spend that effort on the players with the greater ability to produce quality shots.

But what this all means is that I was wrong. Twice. The eye-test failed me on the ice and on the first spreadsheet.

The major underlying issue with Kadri’s line is that they spend too much time in the defensive zone. They’ve been rock-solid while there, but they need an exit strategy, and that definitely was their problem all last year.

Kevin mentioned that they don’t have a real puck carrier/zone entry guy like Nylander or Mitch Marner, and that’s likely what is keeping them from fulfilling their offensive potential.

I still think that their reliance so far on Komarov to be the net-front guy is a problem. Corsica’s Expected Goals do not factor in shooter skill. While, so far, Kadri’s line have outperformed their expected goals in actual scoring by a tiny amount, we should ask how much more could they do with Kadri and Marleau producing better quality shots.

They need to be shooting more than league average quality and in locations in line with their historical tendencies, and they likely will with time, but so far, they aren’t doing that. Komarov is the player on the line with the highest individual Expected Goals. Kadri’s is very poor. Which does line up with my view of those shot plots and my eye-test.

The Leafs laid down three fun games that produced three wins and six points, but only one line did high-quality offence, while one did defence, and that has to change. One-line teams don’t win playoff rounds.