3 Myths about the Maple Leafs

Fans always believe in things about their team that likely aren’t true. Most of it is just opinion, or someone clinging onto a not totally impossible idea that’s nonetheless highly unlikely. But sometimes they’re just flat out wrong.  Today I’m poking at three widely stated beliefs about the Leafs.

Myth One

Mike Babcock plays the second power play unit way too much

This is a bit subjective because all you have to do is put “too much” on a sliding scale and it’s always true or false depending on your setting. But for a great deal of the season so far, fans have been seeing instances of the first power play unit (the unit of doom and despair) coming off with almost exactly one minute left in the penalty. The intuitive leap was made that Babcock was playing the two units equally.

This is clear and unambiguous:

The first unit plays approximately 60 per cent of the available time. The second unit plays 40 per cent, and that’s been true all season long.

For all the swapping out of forwards on the second unit the real difference between the two groups is Morgan Rielly. He’s absolutely superb this season at walking the blue line and keeping the puck moving. Next time the Leafs have a power play, watch him, and ignore the Marner, Tavares and Matthews show. He’s so good, he’s making the second unit look bad when they really aren’t.

NHL Defencemen Power Play (minimum 50 min TOI)

Mo!

Rielly leads the NHL in shot rate on the power play, but Jake Gardiner is 20th for defenders. The only team that gets two defenders in ahead of Gardiner is Carolina. Playing that second unit enough to keep them rolling just makes sense from a game-management perspective, but they aren’t being played too much.

Myth Two

Garret Sparks has been good

We’re in the realm of defining good to win the argument either way on this one. My definition is simple: better than most other backups.

For goaltenders, like all players, there’s two bars you can clear: League average, which is just what it sounds like, and replacement level, which is a fuzzier idea, but essentially means the bare minimum of quality on an NHL team.  Replacement level is a small amount below average for goalies in the NHL, where there is a very small clump of very good performers in any one season, a clump around the average, and a very long, dwindling tail through replacement level on down into the depths of truly bad performance. In other words, there’s more bad goalies than good.

League average this year is lower than it has been, and if your mental picture of an average goalie is a medium-quality starter, you’re off the mark. There’s more than 62 goalies in the NHL, and average makes you somewhere in the middle of the clump of backups and failing starters. Please note that all data in this section is all-situations.

The simplest way to address Sparks’ results is to look at Moneypuck’s Save Percentage above Expected. I set it to a minimum of six games played (Sparks has 10) and by this measure there are 13 goalies worse than Sparks and 52 better. Those numbers should tell you that most of the backups in the NHL have performed better by this measure.

Another method, for those who don’t like Expected Goals, is to look at Goals Saved Above Average*. This is a very simple calculation that erases the shot rate differences goalies face. First, the league average number of saves on the total shots they’ve faced is calculated and then the actual saves over or under that is expressed as a raw number. This works great when just looking at starters who have played similar numbers of games, but to make it work for all goalies, rating it out per 60 minutes is necessary.

By this measure, and for goalies who have played at least 6 games, Sparks is 39th out of 71 goalies. He’s the first goalie to be in a negative. So without a shot-quality weighting (Expected Goals is merely a weighting of unblocked shots for quality) Sparks looks like he’s essentially in the  league average group. With the shot-quality weighting, he’s replacement level.

*It has come to my attention that if you’re used to using Corsica for your goalie stats, none of this will make sense. Corsica has redefined GSAA so that the “average” they’re comparing to is their expected goals model. Therefore it’s more like using someone else’s Goals Saved above Expectation. The original form of GSAA, as used by Natural Stat Trick, is exactly as I describe it and does nothing but adjust for shot volume variance and minutes played from one goalie to another.

The reason weighting for shot quality hurts him is that in the 10 games Sparks has played, he’s faced an unusually easy quality of shots against. This is what happens when you look at small samples. It’s not that the results are inaccurate, it’s that the conditions they took place in aren’t necessarily run of the mill. Frederik Andersen faces a lot tougher shot quality on average, and we could expect that if they each played half the season, their experiences would trend toward equality.

One thing that’s a sub-myth of the “he’s good” myth is the idea that Sparks has shown steady improvement. Using Offside Review’s goals against model since it’s got game-by-game data, I came up with this version of events:

This is Expected Goals Against less actual. The higher the positive number the better the goalie did at preventing goals over what was expected of a league average goalie given the shot volume and quality. The rating per 60 minutes accounts for the various lengths of TOI: one game is only 13 minutes, some went to OT.

There’s no trend here. There’s neither growth nor decline. The distribution here is very flat as well. The best you can say is there is only one absolute stinker of a game.

But what about testing the question this way:

Is Sparks good enough?

You might say that a goalie who can hit the mark somewhere between replacement level and average is good enough. I might say that too! He’s not paid much, and for the cap hit spent, he’s likely good value.

Forget averages and expected, and let’s look at real goals for a minute, again rated out per 60 minutes to express each game as a full 60 minutes.

The Leafs score 3.7 goals per game on average. That’s enough to have won five of those games. The actual win/loss rate is very different because in some of these games, they scored more than usual. But on average, the Leafs can win 50 per cent of the games Garret Sparks starts if he keeps on playing like he has been.

So, yeah, that’s likely good enough for a rarely used backup at his cap hit. He’s not been good or above average or even much above replacement level, but he’s good enough for his current usage. But what about the desire to use Frederik Andersen less and the backup more? Sparks needs to be a lot better in the future to make that a good idea.

Ten games is not enough playing time to tell us what to expect in the future. He might get better or he might get worse or he might continue to perform at this average. There is nothing here to indicate one outcome is more probable than the other, so for fans, you can decide to hope if you are inclined that way. The Leafs need to be thinking about what they’ll do if he does not improve or gets worse.

Myth Three

Patrick Marleau just had a slow start and he’s fine now

Now we’re defining fine. How about instead, we compare Marleau’s first 19 games to his subsequent 20 and see what’s different?

Patrick Marleau at 5 on 5

This is not a pretty picture. Marleau’s usage in terms of share of TOI has remained stable, and everything else is worse from last year. In addition, there is no significant improvement from the first half to the second of the games played this season.

His individual shot rate (iCF/60) has climbed a very small and likely meaningless amount, but it’s stalled out lower than his normal amount as seen last year. He’s become a player who shoots at a rate poorer than the all but the worst players on the team. His Expected Goals For (ixGF/60) has climbed a little, so he’s improved in shot quality a touch as well as volume, but it’s dramatically lower than it was last year, and is lower relative to the rest of the team than it was.

In terms of his on-ice performance, he seems to always be a little below the team average in shot share (CF%) and worse in Expected Goals share (xGF%). But in the last 20 games, that last measure has plummeted to “Why are you on the ice?” levels.

The underlying halves of Marleau’s Expected Goals follow a pattern: He’s much better at the for than the against to such an extent that in past seasons and early this year he was very high on the team in Expected Goals For per 60. All his offence was just all wiped out by really poor defensive results. This last 20 games, his xGF/60 has plummeted.

Some things mitigate his performance. He’s played a lot with Auston Matthews whose line has been consistently terrible at shots or expected goals against. Marleau has also played a lot with Nazem Kadri, so a look at their on-ice results together and apart is a little less gloomy:

With and Without at 5 on 5 Score and Venue Adjusted

Maybe just don’t play Marleau with Matthews? They might get broken up permanently soon, so that could help. Meanwhile, Marleau is genuinely better with Kadri, and the offensive pace is excellent when they’re together. Their loopy GF% has led us all to think that they are really out-of-sight good. At least a lot of that turns out to be real. But how much is him, and how much is Kadri?

Marleau’s performance is not bad luck either. His PDO is always above 1 and his usage is not unduly heavy.

If we feel like he’s improved, it’s because of that GF% with Kadri and possibly his power play performance. His crucial underlying numbers are worse in many ways, and nothing spells doom for a top-line forward like a declining personal shot rate. Play him less, in less difficult situations and he’ll likely be an okay player. But the idea that he’s bouncing back is more myth than reality, and concerns over his performance are very justified.