Hockey Reference produces the best visual of a team’s win/loss record. This is the Leafs so far:

It’s obvious that the Leafs started out okay, got stronger, then got really bad, rallied a little, got bad again and are on the up, right? That’s the clear conclusion from this data that’s 100% accurate, and displayed in a way that isn’t misleading.

Not one word of that conclusion is true. And the failure lies, not in the graph nor in the narrative built around some numbers (the bars are just pictures of numbers), but in the assumption that underlies it. Busting that assumption is the reason why so-called advanced statistics came to sports. But even for people who understand shot-based metrics in hockey and who know that plus/minus is “bad”, the assumption persists, so deeply worked into brains that the truth will always get you an argument because it doesn’t feel true. And this is the age of feelings after all.

The big lie of hockey always feels true. Here it is:

Wins and losses and the goals for and against that create them accurately describe team quality and predict future success.

It’s easy to see why that lie is so accepted as true. You have to win to get points, and points in the standings are the marker of success that gets you into the playoffs. If a team never wins and finishes last, they will be obviously bad to any decently experienced observer. The team that wins a lot is some level of good and looks it. Also, watching a team score more goals than the other just makes you feel differently about them to watching one lose.

The truth is that, if all you want is a rough idea of which teams are good and bad, you can go ahead and use points — once there’s a decent number of games played, and if all the teams have played about the same number of games. Be sophisticated and use points percentage, the NHL even provides it now on their standings page. Or once the season is almost done, like now, just use goal differential, it will rarely lead you astray.

Points and goals can give you the broad strokes, but for fine detail, to tell the story of a season, if you use goals, you’ve told people how it felt, and they already know that. What’s the real story?

The Leafs Season

Using Evolving Hockey’s Expected Goals model (all unblocked shots weighted by several factors, primarily location), I looked at a few different ways of showing the whole season, and decided this was the most representative of the full picture.

Normally, we tend to look at five-on-five play in isolation because it’s more predictive, but I’m not predicting anything here, I’m describing, and the special teams play is part of how the team has played. This is muddied by how many opportunities they have for and against on the powerplay, but that can’t be helped.

The line is a rolling five-game average to keep in the up and down swings of a season without being so jittery it stops meaning anything. What this shows is that the Leafs have never been bad. There is one value on that line below zero, which means that the Leafs have been the better team in most of the games. At the game level the picture is a little less rosy, and the line dips below zero more often, but there are still very few games with a negative number.

If you want to play with more views and with things other than Expected Goals, Moneypuck’s interactive graphs are for you. Their Expected Goals model is a little more exuberant in how it measures the power play shots, but the overall concept is the same.

One caveat on a rolling average like I used is that, while it blunts the edges of extreme outlier results, it doesn’t remove them, so I checked the game level on Moneypuck to see if there is a situation like we had last year I needed to address. Last season, the Leafs played the Red Wings in a game shortly after Keefe took over, and they so overwhelmingly overwhelmed Detroit, the data from that game skewed the averages to such a degree that I thought it was more accurate to leave that game out.

This year there is a game on January 18 vs Winnipeg that is yanking the early portion of that line above zero all by itself. It’s a level of dominance the Leafs have repeated and bettered since, so it’s not quite an outlier, but it helps undersell how poorly the Leafs performed for the first few weeks of the season. The drawback of looking at an average is that we naturally imagine the team performed at that average, when that is rarely true.

The Leafs did perform poorly at first, relative to our reasonable expectations for a team expected to dominate the division. Every recap or article early on  that discussed the Leafs poor five-on-five performance, and the power play success that was hiding it, resulted in some comment quoting the win-loss record. Or, in other words, an expression of frustration that the analysis didn’t match how it felt to watch the goals for and against and the team win.

Since that period, the five-on-five play has settled into a reasonable fluctuation (based in large part on who the opposition is) that averages out to good to excellent through the most recent three months of the season. The power play has looked bad, felt frustrating, and has failed to produce goals, but hasn’t been terrible in a real sense, just below our expectations in the same way the five-on-five play was early on.

There hasn’t been much point in deep dives into the Maple Leafs play because it’s been boringly good, boringly consistent, and there’s just not anything really wrong with their performance. Sure players made mistakes, and the game of fixing blame for every goal against while giving full credit to the forwards for every goal for creates exactly the expected negativity about every defenceman or pass-first forward, but the team has been very good, shading to excellent for three months.

It just hasn’t felt that way lately.

Goal differential in a game is made up of the way the team plays — the Expected Goals describes that decently well — plus the goaltending, the shooting skill of the players taking the shots, and random variance in all of the above plus a few dozen other things. When you see a goal differential diverge from the Expected Goals so dramatically, it might not be just the goalies.

It’s Just the Goalies

That line, which very nearly matches the goal differential line, is all four goalies measured game by game in Goals Saved above Expected, and then averaged in the same way the Expected Goals and actual goals were. If the goalies were having a collective league average sort of season, that line would match the Expected Goals more. Since it’s closer to the outcome — the goals +/-, it’s safe to say it’s the goalie performance driving that outcome away from the expected results.

Now that I’ve looked at all these lines and numbers, I can tell a story about the Leafs season more true than that first one:

The Leafs began with okay goaltending, troubling five-on-five play and a power play so amazing, it covered over all other flaws and banked some wins that came in very handy later. The goaltending got worse, and hasn’t ever been good enough, and the five-on-five play soared to the level we really expect from the Leafs, at times disguising the bad goalies and making up for the ever deteriorating power play.

The team is fine, great even. The defencemen aren’t terrible, they don’t need to be benched and the forwards are not ruining the team by being old and slow. The Leafs aren’t doing better because they’re tougher now, and they aren’t going to go on a losing streak and not make the playoffs. The most important part of a team’s future success — the five-on-five play — is what’s working consistently well.

There’s just no other narrative that fits the facts. The questions for the playoffs are the goalie and the power play in that order of importance.

The trouble is that even a full ‘go for the cup’ playoff run can be a run of play so short, results can be skewed dramatically by random events.  Random is absolutely code for goaltending in that sentence because the most variable result on most teams is the goalie — unless they’re consistently bad, where you get to a point where how bad becomes irrelevant.

The job of the GM, the coach and the players is to control what they can control and forget the rest. The Leafs have produced a game that touches greatness at times, and is nearly always very good. That’s all they can do, the rest is up to fate.

Analysis