clock menu more-arrow no yes

Filed under:

PD Oh No: The Maple Leafs, Percentages and You

It’ll be fine. Honestly. It’s just randomness.

Toronto Maple Leafs v Buffalo Sabres Photo by Ben Green/NHLI via Getty Images

You like Michael Bunting, I like him, and he’s fun. He skates around chewing his mouth guard, gets into chirp-fests with the rat one bench over (particularly when that’s Travis Konecny), and he drives the net like it’s his job. That’s because it is.

But is Micheal Bunting a skilled shooter? How can we tell, and more to the point: So what if he isn’t? He’s in the crease to collect garbage goals like Zach Hyman used to do. Bunting does his job very well, and no one wants him to go anywhere, but he’s one of the group of Maple Leafs who seem to not be scoring much so far. The team has 38 Expected Goals at five-on-five and have actually scored 28. That’s going to change, though, right?

If you check Twitter, the answer to all of this is so simple. It’s just PDO. And the way to look smart, as many, many people have said over the years, is to predict that your team will start winning more when their PDO is bad and everything looks bleak. If what you’re after is looking clever, that’s as far as you need to go, but if you want a finer granularity on your understanding of shooting, skill, luck and happenstance, then PDO is not enough.

PDO

Now, before I go back to ignoring it forever, what is PDO?

The name means nothing, which is apt. The stat was invented back before the earth cooled, and PDO simply takes old-school shooting and save percentages and adds them together. It was the bees knees ten years ago:

That article quickly sums up PDO as useful because the underlying measures in it are “primarily luck driven”. And my response is to ask: Are they, though? To settle that would devolve into a fight about the meaning of primarily, so I’ll pass, but my take on PDO is this: both shooting % and save % are made up of individual and team skill, competition, coaching, luck, and potentially other measurable factors we haven’t figured out how to quantify yet. And this is the key: in no identifiable proportion.

PDO looks very meaningful because at the league season level, every shot on goal appears in both figures. Every goal lowers a save % and raises a sh% and every save does the opposite. So if you add up everyone’s %s at the end of the season you get a single answer every time: one.

This simple reality has led to a mental framework that says at the team level, high PDO will regress back to one and low will regress up to one. The same idea is applied to the individual. But the truth is, each team and player won’t end up with a PDO of one, they will land where they land. So, the mental model gets adjusted to PDO that’s really extreme will regress, but there is no real line demarcating the muddy middle from the extremes.

Back in the last normal season, 2018-2019, the Capitals finished with a PDO of 1.022 and the Wild had .984, marking the bounds of that season’s range. This year, with only a few games played by each team, the Capitals lead with 1.038 and Chicago are at the bottom with .947. It’s not all that likely those two teams will stay there for a whole season, but it is wrong to assume very good or very lucky teams can’t sustain values well over 1 for a whole season. Very bad or very unlucky teams can stay in the basement all year too. The prophet using PDO to look smart needs to confine himself to the extremes.

A glance at PDO can’t tell you a thing that makes you smarter. Are the Caps shooting hot, or is it the goalies? Both? Is it luck or Ovi? We’re all in the dark here with this miracle number. It’s not actually more useful than looking at the standings.

I said there’s no known proportion of the ingredients in shooting and save percentage, but it is very clear that their recipes are broadly different at the team level. A team has one or two goalies who play most of the games, and they have more clearly understood ranges of results. The dozen or two skaters who may play in a season are a much more diverse group, and should more likely trend to league average in results over enough time. This is why adding these numbers together and expecting to get smarter for it is just not going to work. Mixing Jack Campbell’s start to the season with all of the Maple Leafs skaters and the few games played by other goalies, gives you a stew with too many ingredients to name from one taste. This % gumbo needs to be broken down into smaller segments and examined to see if the Maple Leafs are just unlucky or if there’s something else at play.

What Can We Expect?

With PDO thrown out the window, shots on goal can sail out right behind it. Reminder: throwing terrible stats out of your window is polluting, and you should dispose of them properly. I’m not interested in goalies right now, so they can go away too. I want to look at the Maple Leafs and their shooting. This means that horrible acronym is coming out to play: xFSh% or Expected Fenwick Shooting %.

  1. Fenwick is all unblocked shots, so that’s goals, saves and misses.
  2. Expected goals models are all based on Fenwick because the NHL records blocked shots at the location of the block, rendering the location data useless.
  3. Expected Goals is Fenwick weighted for location, shot type and some other factors and tells you the chance a league average shooter will score with that shot.
  4. Fenwick Shooting % is the percentage of goals scored out of all the unblocked shots taken.
  5. Expected Fenwick Shooting % takes Expected Goals and calculates what a league average player should get in Fenwick Shooting % out of all their unblocked shots.
  6. The point of all this is to compare the Expected to the Actual, and see who is getting more or less goals than an average player would likely get taking the shots they have taken.
  7. Let’s do that.

Maple Leafs as of November 15 at five-on-five only

Player Fenwick Sh% Expected Fenwick Sh% Difference
Player Fenwick Sh% Expected Fenwick Sh% Difference
Michael Bunting 7.5 11.12 -3.62
Wayne Simmonds 3.23 9.45 -6.22
Nick Ritchie 0 9.45 -9.45
John Tavares 11.63 8.88 2.75
David Kampf 13.33 8.33 5
Ondrej Kase 8.51 8.02 0.49
William Nylander 5.45 6.71 -1.26
Alex Kerfoot 6.45 6.48 -0.03
Mitch Marner 5.13 6.23 -1.1
Auston Matthews 3.28 6.21 -2.93
Travis Dermott 0 5.5 -5.5
Jason Spezza 7.41 5.11 2.3
Pierre Engvall 0 4.18 -4.18
Rasmus Sandin 0 3.33 -3.33
Timothy Liljegren 0 3.11 -3.11
Morgan Rielly 2.86 2.63 0.23
Jake Muzzin 3.85 2.04 1.81
Tj Brodie 0 1.86 -1.86
Justin Holl 0 1.75 -1.75

Note: you can look at actual goals over Individual Expected Goals at the skater level to see a slightly different view of the same information. That information inherently contains shot volume which is the largest component of Expected Goals. Sh% can more clearly show you shooting talent/luck because it’s not influenced by quantity.

Let’s take an easy example here. Auston Matthews is not going to spend a season shooting almost three percentage points below what an average shooter should accomplish. Since 2019, so covering the team under Sheldon Keefe’s system, Matthews has shot almost four percentage points over expected. Predicting he’ll score more in future games will make you a prophet. How many and when? That’s not predictable at all.

Michael Bunting leads the Maple Leafs in Expected Sh% with 11. In fact, that number is the highest of any player in my “since 2019” list. It’s not sustainable in the long term, but it’s not unreasonable that a player with his net-driving ways has a high expected value. Shots close to the net have a higher chance of becoming goals. But will his actual goals rise? We have no way of guessing. He has too little time in the NHL to give us any information on him, but watching him play leaves you feeling like you’re seeing an opportunist who makes up for his lack of shooting skill with a willingness to always make more opportunities. He may stay with a negative differential for his whole career, we can’t know.

Nick Ritchie, Wayne Simmonds and David Kämpf are all very high in expected, and have varying actual results. They all have some similarities to Bunting, one of which is that they don’t seem to be quality shooters. Nick Ritchie will score a goal eventually, Wayne Simmonds will not score like he used to, and Kämpf is not going to have a Fenwick Sh% of 13 all season without a lot of luck landing in his lap. Luckily, he’s on a two-year contract, so no one will be paying him for a lucky shooting season.

Alexander Kerfoot is having a weird season. You might guess he’s not a very offensively gifted player, and you’re mostly right, but he does what Tyler Bozak used to do, and he is selective in when he shoots. Usually his actual percentage is very high, and we might see some more goals from him.

William Nylander shooting right around 8% in Fenwick money, and right around Expected is normal for him, so like Matthews, he is not quite getting the quality we might expect. John Tavares is, and maybe we shouldn’t count on his scoring continuing.

What does this all add up to? Well, a big confused picture of early season results, that’s what. If the current structure of some of the best players getting Expected Shooting %s below their norms and shooting under the expected on top of that continues, then the Leafs may well continue to score fewer than expected as a team. Meanwhile, don’t look for the top three in Expected Sh% to ride to the rescue.

PDO is a broad strokes measure that is good for arguing with someone using plus/minus or GF% (plus/minus in a flashy suit) to say the Leafs are bad, har har. Looking at individual results over expected is better, but each of their results are their own gumbo of various factors, and you shouldn’t expect every player to shoot over expected. The debate naturally tends to the extremes when looking at shooting percentages. The Leafs aren’t bad, says the sage, just unlucky. Okay, that seems true, but how good are they going to be when those percentages settle down over the season?

That’s the really tough question, but this data has planted some seeds of unease in my mind, even as I know full well Matthews will score more. His expected % shouldn’t be in the bottom third of the forwards. Neither should Nylander’s. If the most highly skilled players are not the ones producing all those shots creating that big Expected Goals number, if in fact a lot of below average players are, then can you expect the Leafs to score above their expected? Keefe has taken the puck away from the defence a lot, and you can see why that’s good in those numbers above. But given the roster as it exists, he’s given it to forwards who might not actually be any more likely to score than Jake Muzzin.

A fight that asks are the Leafs good or bad is pointless. When the question becomes one of what flavour of good the team is, one thing stands out: The Leafs lead the NHL in Expected Goals For and their offence is excellent in theory. Can we expect the reality to match that theory, or are the Leafs a team that will continue to underperform their expected goals?

Time will tell us.