Staturday Weekly Column #8: Corsi Report Cards!

0
12



As the NHL season quickly progresses, we find ourselves already at game #12 tonight for the Maple Leafs, and we are just crossing the precipice of having enough data to semi-accurately predict future results, at least when it comes to the basic counting stats.

Research has shown that Corsi can predict the win % of teams, using even a pretty small sample such as 20 games played. The key point from the linked article from JLikens in the early-ish days of blogging about shot attempt analysis (2011) is the following two points:

  1. Corsi Tied is the best predictor of how a team will perform over the remainder of its schedule, regardless of the point in the schedule at which the calculation occurs.
  2. Corsi Tied is only marginally more predictive of future success than goal ratio or winning percentage when looking at samples of 60 games or more.

So, in the early parts of the season like right now, Corsi is at it’s most valuable in terms of predictivity relative to more basic metrics like Win % or goal differential.

Advertisement – Continue Reading Below

The above article talks about “Corsi Tied”, which was an early, simple way of eliminating what we call “score effects”. Score effects are basically: as the team who is winning gets closer to the end of the game, for whatever reason, they typically allow more shot attempts. This is proportional to how much time is left, and what the score differential is. At the same time, we knew that Home teams were generally favoured over Away teams. Using these, more modern analysts have been able to create a Score and Venue Adjusted Fenwick model, and Score and Venue Adjusted Corsi model, which instead of eliminating all events when the score isn’t tied, instead adjusts the shot attempts to be less valuable or more valuable depending on the score and the time left. This kind of data is what we’ll use today.

Some have stated that Corsi isn’t all that valuable anymore as teams have learned more about this data and started to build their teams around this kind of analysis. However, it still has some predictive value, and that’s why I’m still writing about it.

On to what we came here for, which is to answer the question of how each Leaf has been doing this year in terms of shot attempt metrics.

We’re going to use a total of six numbers to show this: score-adjusted CF%, xGF%, GF%, and PDO. Basically, CF% is shot attempts, xGF% is shot attempts that we’re applying a location adjustment to, for both of which we’re applying a score and venue adjustment to as discussed above. GF% is on-ice scoring, like +/- but more refined, and PDO is luck. Check out Part 1 of my advanced stats primer to learn more about what these are. The last two I didn’t list, which are WAR and xWAR, or Wins Above Replacement and Expected Wins Above Replacement. These are a relatively simple concept: WAR is how much better your performance has been from what we’d expect from a “Replacement level” player; and xWAR is how much we’d expect their performance to be in the future in terms of WAR. The work behind these are a little more nebulous, so read Part 2 of my advanced stats primer to learn more about them.

Advertisement – Continue Reading Below

Non-WAR will be at 5-on-5 and will come from Natural Stat Trick. WAR includes all types of states, including powerplay and shorthanded rated differently than 5-on-5, and it comes from Evolving Hockey.

The Team

Starting with the team as a whole, we have the following stats:

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO
53.18% (6th) 50.95% (15th) 56.86% (6th) 1.016 (9th)

Looking at all this, we see that the Leafs are one of the best teams in the league at flinging pucks towards the net, and also one of the best teams at getting those pucks to actually go in the net.

Advertisement – Continue Reading Below

When we adjust for the location of those attempts though, using xGF%, we find that they’re a middling team, which is uninspiring to say the least.

Their PDO, or luck factor, also sits a little high (>1 means you’re lucky, <1 means you’re unlucky).

We can expect, then that the goals to dry up a bit, but if you trust Corsi more than xGF% like friend of the blog draglikepull does, the Leafs should be in good shape for the remainder of the season.

Advertisement – Continue Reading Below

I framed all of the above from an offensive perspective, but it’s important to note that each of these statistics includes the defensive side of the puck as well.

What about report card grades? I’ll give the Leafs as a whole a solid B+, as the CF% is really encouraging, and I’m not that concerned about the PDO since the Leafs generally are a high shooting percentage team sunk by bad goaltending, but that goaltending has been turning for the better lately.

The Skaters

For the Skaters, we’ll sort by time on ice, only including those players who have at least 100 minutes of time on ice so far. This basically filters out the 4th line players who we don’t have nearly enough data on yet to see how they’re doing this season (except hat-trick-scoring Jason Spezza, we don’t need data to know he’s doing great).

Basically, at the top of the list you should be drawing the strongest conclusions, and the bottom of the list your conclusions should be taken with a boulder-sized grain of salt.

Morgan Rielly

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
48.93% (11th) 47.33% (9th) 51.42% 1.018 0.2 0.1

The star defender for the Leafs has been less than impressive this year in terms of these kinds of numbers. Of 13 skaters that meet the thresholds, he’s near the bottom of the team. He’s also had pretty high PDO and not all that impressive of a production spell, so things could get even worse for Rielly. Ultimately the Leafs need more than replacement-level hockey from Rielly, but xWAR tells us that that’s basically all we should expect.

Grade: D+

TJ Brodie

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
48.84% (12th) 45.11% (12th) 51.05% 1.014 -0.2 0

It would be really, really weird for Brodie to be very far from his partner Morgan Rielly, since the majority of these stats are “on ice” and not “individual” stats. Brodie somehow rates even worse than Rielly, actually bringing a negative impact so far in terms of WAR. On the bright side, xWAR tells us this will improve slightly, but they will definitely want to see more from their new signing playing on the top pairing.

Grade: D

Jake Muzzin

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
55.08% (5th) 53.60% (6th) 60.23% 1.017 0.2 0

Muzzin on the second pair has seemed quite good. He’s done great in both Corsi and xGF%, and his GF% is high. That production will go down a bit with a slightly high PDO, but it’s starting high, so that’s good. In WAR though, he’s just been okay, and xWAR tells us he’ll play like basically a replacement player in the future, which is certainly concerning.

Grade: C+

Justin Holl

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
57.43% (2nd) 62.35% (1st) 64.73% 1.013 0.8 0.7

How about Justin Holl, eh? After many complained about his presence early on, he’s become the Leafs’ best defender in these early days. All of his numbers are incredible, and he’s even got the lowest PDO of the top 4 defenders. What else can you say but good for him.

Grade: A+

Mitch Marner

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
52.64% (9th) 50.07% (8th) 59.55% 1.035 0.3 0.2

It’s hard not to expect the world of Mitch Marner given his contract, but even acknowledging that these numbers are a bit disappointing. He’s got a sky-high PDO, and average-at-best xGF% and WAR numbers. His Corsi is good in that it’s a healthy amount above 50%, but being 9th of 13 skaters for Toronto is disappointing.

Grade: C-

Auston Matthews

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
56.60% (3rd) 55.37% (4th) 59.83% 1.022 0.7 1.1

Matthews, while being a linemate of Marner’s, has fared far better somehow. His WAR numbers are near the top of the team, and his xWAR is actually at the top of the league. Both his CF% and xGF% are excellent, and his PDO is high but not so high that it’s going to crash like the 2008 recession.

Grade: A+

John Tavares

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
54.47% (6th) 54.70% (5th) 59.84% 1.018 0.2 0.3

The Leafs are blessed to have two elite centers, but only one is playing truly elite right now. Tavares has been good, don’t get me wrong, but he’s not at that elite level right now. The WAR and xWAR are not great, but the CF% and xGF% are good, so he’s not in a concerning place at all. Overall I’m not concerned, but it should be noted that he hasn’t been elite yet this season.

Grade: B

Zach Hyman

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
53.05% (8th) 56.24% (3rd) 63.29% 1.041 0.2 -0.1

Hyman is the work horse top-6 forward that you don’t necessarily expect to be great at scoring, but you do expect him to be a good influence on the shot attempt numbers by forechecking and backchecking strongly. He’s been good, especially in the xGF% department, but the xWAR is a bit concerning. His GF% is very high, and the PDO is atop the team, so be wary that the scoring numbers while Hyman is on the ice could drop in the near future.

Grade: B-

William Nylander

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
58.85% (1st) 56.99% (2nd) 63.08% 1.026 0.3 0.3

Nylander is dominating the Corsi and xGF% numbers this year, but that hasn’t quite shown itself in the WAR numbers totally. He’s got a high PDO and a really high GF%, but the shot attempts are flying around Nylander. He’s been a great passer setting up Tavares and others, and hopefully the more nebulous stuff starts to turn in his favour as well.

Grade: A-

Ilya Mikheyev

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
48.84% (13th) 45.63% (11th) 40.99% 0.985 -0.2 -0.1

The Leafs’ soup-loving sophomore is having a tough start to this year, after such an incredible start to last year. He’s at the bottom of the team in CF%, and looking poor in terms of WAR and xWAR. The goals have not gone his way either, but that should change any time now with such a low PDO.

Grade: D+

Zach Bogosian

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
53.76% (7th) 46.41% (10th) 54.88% 1.009 0.1 -0.1

Bogosian is one of those much-maligned players who never get the benefit of the doubt from nerds like myself, but it’s easy to say he’s been fine. There’s nothing here to say he’s been great, and that xGF% crater is quite concerning, but for the role he plays he’s doing ok.

Grade: C+

Jimmy Vesey

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
55.93% (4th) 51.08% (7th) 49.86% 0.976 0.4 0.4

Jimmy Vesey is another one of those player types that gets unfair stick, but for different reasons. He’s more like a Marincin type, who has size but doesn’t appear to use it, and enough skill to be dangerous to the opponent but that creativity puts him in danger from the opponent as well when it goes wrong. He’s been good, though much of that has to be playing the majority of his time with Tavares and Nylander in a top-6 role that he hasn’t perhaps earned.  The numbers look good though, and I’m not concerned about his role at all.

Grade: B+

Alexander Kerfoot

5v5 Score/Venue Adj. CF% 5v5 Score/Venue Adj. xGF% 5v5 GF% 5v5 PDO WAR xWAR
49.54% (10th) 41.39% (13th) 38.14% 0.978 -0.1 -0.2

Kerfoot has taken a lot of abuse in Leafs fandom for being poor this season, and in this case the numbers agree. The production has been poor, the shot attempts have been poor, and the WAR has been poor. He’s not been good enough, at all.

Grade: F

In summary, the Leafs get the following grades:

  • TEAM: B+
  • Rielly: D+
  • Brodie: D
  • Muzzin: C+
  • Holl: A+
  • Marner: C-
  • Matthews: A+
  • Tavares: B
  • Hyman: B-
  • Nylander: A-
  • Mikheyev: D+
  • Bogosian: C+
  • Vesey: B+
  • Kerfoot: F

The positive surprises from Vesey and Holl outweigh the negative surprises from Brodie and Kerfoot, but there’s room to improve all around. We’ll do something like this again around the halfway mark of the season, to see what’s changed, and because at that point the data will be at its most predictive in this shortened COVID season. That of course assumes we get to the halfway mark of the season…

Anyway, stay safe, wash your hands, don’t travel, wear a mask, and Go Leafs Go!