A Tautology of hockey, or why the Leafs blow leads

Play Right

Mike Babcock changed the slogan on the Leafs dressing room wall recently to Play Fast. Play Right.  As opposed to play fast and loose, one imagines.

I confess that I am one of those people who roll their eyes and make jokes at slogans.  I think demotivational posters are funny, and I am not the sort of person who longs to have someone give me a mission statement.  I am not the personality type of an NHL player.

This change in slogan happened in the reasonable temporal vicinity of some poorer play by the Leafs, a downturn in wins, and some games that were lost after the Leafs had been leading.  It seems, once I’ve stopped rolling my eyes, to be fairly good advice.  There is the ghost of a tautology in there: play right.  As in: play correctly, as in: play better than you have been, as in: if you play better, you won’t be as bad.  Well, you won’t, will you?

I Don’t Care That it Was 4-1

At approximately the same time as that was going on, Hockey Graphs published an article by Alex Novet called There’s No Secret to Protecting a Lead.  And I confess again that I grabbed onto that with the warm happy glow of having my biases confirmed and made it my credo.

The gist of the hockey graphs article is another tautology.  You give up goals when you have the lead because you give up goals.  Or: you’re bad in the third period in the same ways you are bad all game.

This is not something people want to accept. We see the blown lead—and that very term is laden with assumed significance and emotion—and we know it matters.  It isn’t at all the same as giving up an early goal or a second period goal or a goal that makes a 4-0 game end at 4-1. It’s a disaster.  It’s a rug pulled out.  It’s a bait and switch. You made me love you, and then look what you did!

And then to make this all more complicated, we thwarted lovers take our human brains that evolved to make patterns out of half seen shapes and draw conclusions from correlations, and we connect the dots of this blown lead, that one, the one before and a pattern emerges.  There has to be one.  We know it. We feel it.

There is a logical explanation for this.  Forget love or betrayal, let’s look at this systematically.

Back to Hockey Graphs.  He looks at three specific types of lead protection.  All of it lumped into one big pile, so any lead of any amount at any time, as well as leads in the third period, and leads in the final five minutes of games.   This is, to be clear, a lot of data over a lot of years, not just a few examples from one team strung on a string.

The average team will have an LPR [lead protection rate] of 43%, meaning that 43% of the time that they get a lead, they will hold on to that lead and win the game. Unsurprisingly, when you narrow the view to leads later in the game, the protection success rate goes up; the average LPR-3 is 59% while the average LPR-Late is 89%.

So we begin with an actual measurement of how often an average team holds a lead. Not all teams are average, of course.  But is this measurement of how a team protects a lead of any use?  If I sit and calculate those rates for the Leafs this season, do I then know the Leafs better?

Mr Novet digs into that in a very meaningful way, delving into first the repeatability of these three types of lead protection measures and then asking if they tell you if a team is likely to win or not.  Even a layman can get a lot out of his full article, so go read it!  I’ll wait.

What he discovered is that the late lead protection measure is not repeatable, but the other two are.  But along the way, he found that they correlate very strongly with a team’s Goals For percentage.  More strongly than Corsi correlates with GF%.

The correlation of CF% to future GF% (stronger than GF% itself) is the underpinning of the idea that looking at all shots, not just some of them, can tell us if a team will be more or less likely to score more goals in the future, and therefore win more games.  We sit on top of that mathematical truth every day whenever we talk about “analytics” in hockey.  The more obvious truth that goes hand in hand with it is that GF% correlates to winning.  A tautology again: If you score more goals than your opponents, you win games.

If lead protection is even more strongly correlated with GF%, the next question is does it predict anything about a team that GF% doesn’t?

His answer was that no, knowing if a team does well or poorly at lead protection doesn’t help you figure out if they will win more or less games and get points.

He says this:

My key point is that we can measure how well teams have protected leads in the past, but there is no “protecting a lead” skill or system for a team that is separate from their general ability to outscore their opponent in any other situation. Good teams protect leads because they tend to score more than their opponents all the time. Bad teams tend to blow leads because they always tend to get scored on.

Tautology again, if you give up goals, it’s because you get scored on a lot.

But what, you are asking now—or at least I hope you are, because I was—what about scoring effects?

Turtles and Shells and Coaching Decisions

We know about scoring effects, and we know that adjusting our Corsi for scoring effects makes it a better predictor of future success.  This is not a huge improvement over raw numbers! I think sometimes when we say this measure is more accurate than that, we infer massive levels of improvement.  Like with the New and Improved! laundry soap, we should be a little cautious about emphasizing how much better that is.

Nonetheless, scoring effects are real. When you have the lead, you take a smaller share of the shots, and when you are behind you shoot the puck more, and is this psychological, a confluence of coaching between the two teams? Who can say? The answer is likely some of this, some of that.  It happens to almost all teams all the time to varying degrees.

So given that, can you not change your fate? Surely scoring effects are not like a horoscope and beyond your ability to influence?  And if you can influence your scoring effects, you can stop blowing those leads!

Or can you?

The relationship between Corsi and goal scoring is not that direct or immediate.  It’s only part of the story on why teams score goals and win games.  But we seem, it seems to me, to have decided that a “defensive shell” is always bad, that turtling is wrong, and the only way to win a game when it’s 4-1 is to go out there and try to score more, to be just as good or as bad as you always are. To play right.

Mr Novet says this:

Whether or not the team protects leads should not be used as a diagnostic tool. If a team is blowing a lot of their leads, the solution isn’t necessarily to get more shutdown players who can prevent the next goal; rather it might be to get better offensive players so that the lead becomes larger. The route you take doesn’t matter as long as you build a team that consistently outscores their opponents.

So the turtle isn’t really the villain of the piece.  Nor is he the hero.  The implication of Novet’s idea here, to me, seems to be that a team must just be who they are all the time as best they can. For the Leafs then, the turtle shell is likely not a look they can pull off.

But I’ve been puzzling over this conclusion for a while and still saying, “But scoring effects!” a lot in my mind.  You see, I don’t take well to credos, and I keep questioning them.

Habits are Hard to Break

The answer might well be in the older work by Micah McCurdy linked above where he shows that adjusting for scoring effects is better than using an older, popular for a time method called Corsi Close.  He feels very strongly on this topic:

Finally, and least obviously, we see that score-close possession metrics are utterly indefensible for any purpose at any time. Raw measures are preferable for conceptual clarity and for predictivity at almost all sample sizes, and adjusted measures are superior for predictivity at all sample sizes. It is difficult to overstate how important it is that they be purged from the lexicon of all right-thinking people. They purport to distill the essence of possession when in fact they do great violence to data by censoring large tracts of meaningful information and magnifying a smallish portion. Adjusted measures, by contrast, apply small nudges to the raw data—their seeming complexity masks how much closer to raw data they are than ‘close’ measures.

This is the very important point that is missed when fans look at shot data for only parts of games.  You cannot slice out part of a game, magnify it, and gain insight beyond what the score adjusted five-on-five numbers all season have shown you.  We learned this lesson once before.

When Novet says don’t use lead protection as a diagnostic tool, I think he’s right on the money.  I don’t think it’s wrong exactly to examine coaching or player decisions in specific situations, but don’t forget to make sure that what you’re looking at is really different from the range of usual behaviors shown all game long.  Your situational problem may well turn out to be systemic when you back up far enough to see the whole system.

Is the turtle you think you see really a defensive tactic, or is it the inevitable result of the other team dropping their reticence around risk taking and changing their actions?  Why not both, as well as other factors we don’t even understand.  But no matter what, you aren’t alone on the ice, no matter how attractive the idea is that you absolutely control your own fate.

I think personnel deployment is largely a red herring in this quest to stop blowing leads. “It’s all Hunlak’s fault” isn’t just tiresome, it’s meaningless.  There is simply not that much leeway in coaching deployment to keep all the players you’ve decided are at fault off the ice in a given situation.  A short bench is not sustainable. The tautology there is: if only we had better players, we’d be a better team.

So how much can addressing coaching or player behaviour in one game state really change game outcomes? Is Novet right, like my gut tells me he is—and his math, don’t forget—that the answer really is the most painful truth of all: It didn’t stay 4-1 because the team was bad and they were lucky to have got those four goals in the first place.  Ouch.

Tautology Again

Of all of these truisms and obvious things, the one that does stand out to me is this one: If only we were better, we’d lose less.

“Play right” now seems to be advice to just be as good as you can be all the time.  “Play fast” seems to be saying to just give yourself over to who you are as a team.  And it’s not like fine tuning adjustments of situational behaviour don’t matter; they do. Until they don’t.  They matter until they are overtaken, as the math says they will be, by a team’s true nature.  Nature will out.

The solution, after all of that, is to pick a path through the forest to being better and make it happen.  But don’t thread a bunch of blown leads on a string and give it meaning beyond any other small sample of game time sliced off from the whole and examined under a microscope.  You risk failing to see the forest for the trees.

Why do the Leafs blow leads? Because they have the fifth worst goals against rate in the NHL. They can either fix that, or they can recognize that blowing a tire is sometimes the price of trying to outrun your weaknesses with nothing but speed.