Lucky thirteen: A numerical picture of the Leafs first few games

Please note: the numbers and rankings in this post are valid as of Tuesday’s games and do not include Wednesday’s events.

Descriptive vs Predictive

This is an older chart from HockeyViz, that was made to look at which flavour of Corsi best predicts future goal scoring.  The ultimate answer was Corsi adjusted for scoring effects and other factors gives you the best prediction.  This was the old score adjusted vs Corsi close debate.

But the secondary point of this chart is that it shows when that predictive power peaks.  It is not at 13 games.  But 13 games in is about as good as halfway through the season.  We need to be patient and wait for 20-30 games to really get to the meaningful predictive value of shot differential.

Note: the line on this graph drops off very fast because the fewer games left to be played, the harder the outcomes are to predict with any method.  The reason for this is simple.  Hockey is full of random events. The puck bounces how it will sometimes, and not everything that happens on the ice is under your control. The fewer games left, the more that randomness takes over as well as other factors like quality of competition that tend to smooth out over a whole season.

In the meantime, we can look at what happened in the Leafs 13 games to better understand what we saw, and then we can ask ourselves where the team could be if they lay down ten more games of similar overall quality. That’s a big if.  Anything can happen—including improvement, it should be said.

Shots Lead to Goals and Goals Lead to Wins

That seems obvious, but while the chart above shows that shot differentials can predict the probability of future goals if you use the information correctly, other work has shown that goal differential is a very weak predictor of future success.

However, within a sample of games, the strongest link is not shots to goals; it is goals and wins. That might be a rather obvious statement too. Goal differential is the definition of a win in hockey.  Within a sample of games, the relationship between shot differential to goal differential, and therefore wins, is fairly weak.  Much weaker than the relationship to the future that is graphed above.  So winning the Corsi battle in a given game is not a strong indicator that you “should” get more goals and win that game.  Or 13 games.

The other issue when looking at shots and goals is shooting percentage. It varies. And it is understood that it is not a repeatable skill.  That doesn’t mean you can’t make your shot better and improve your game, nor does it mean that one player isn’t better than another at scoring goals. What it does mean is that within a sample of games, you cannot expect the shooting percentages to stay that way in the future. It tends to settle closer to league average for nearly all players in the NHL as a season goes on.  The exceptions to that are a very, very short list. But remember, there is no timetable for this. Regression might be inevitable, but that does not mean it is imminent.

Lucky 13!

With that set of relationships in mind, let’s get to the games and figure out what the Leafs did in them.

Mike Babcock’s Metric

Just for fun, have a look at how Babcock talks about the Leafs performance.  He likes to slice things up into five game segments and he wants six points in each. I set it as a rolling sum of the five games to see how often the Leafs make it to six.

They’re better than last year! But have only hit the number six in the last few games.  They can actually lose their game on Friday and stay at six.  Any point in that game will put them above it for the first time.

Take notice on last year’s line where there is a dip back to zero at game 43 and then abysmal results followed.  That was right after I’d predicted the Leafs could be sniffing at bubble team territory if they kept on the way they were. They didn’t keep on.  Probability is not destiny, and there is no such thing as “on pace for” in hockey.

The Standings

I have a particular way of looking at the standings, perhaps even peculiar.  I go to NHL.com/standings and I set it to League and I click on ROW twice to look at the bad teams.  ROW is regulation and overtime wins.  This is the games you won the hard way and is a more accurate reflection of a team’s play than points.

The Leafs are tenth. That is tenth worst of course, and they have five.  But the number of teams with the same amount is seven, and there are 21 teams with four, five or six. So it is, for real, too early to learn anything there.  The Leafs are in the large group of teams that sit below the fast starters which are largely good teams.

Next, I do the same thing with goal differential.  The Leafs are fifth worst with -11 and that puts them firmly into the bad team group.  It’s not unheard of for teams to make the playoffs with a negative differential, but not one that big.  This isn’t news; there have been two blowouts and some bad goaltending.  Note that the Dallas Stars are in the same situation with an ROW of 4 and a differential of -13 after getting thumped by the Jets on Tuesday.  Remember this for later.

Shots of All Sorts

Using Natural Stat Trick’s Score and Venue Adjusted team data for the season so far, we can see where the Leafs fit in.

In Corsi For percentage, the Leafs are ninth. This is the number you can look at and say: hold that position through the next ten games, and the team has a high probability of a good goal differential and lots more points in the standings. Perhaps many more than have initially been predicted. We are talking about P-word territory here.

We know they don’t have the greatest goal differential right now, and you may have heard that Frederik Andersen played very poorly in some early games.  There was some talk about that.  And while that’s very true, a deeper dive into the shot numbers shows some other weaknesses in the Leafs game.

In Corsi For per 60 minutes, the team is first. The only other team who is even approaching that level of shots raining down on the opposing team is the Los Angeles Kings (to the surprise of no one after the last game).  Most of the teams in the top ten are top-ranked teams, but the Hurricanes are in there too, and the Blue Jackets aren’t far off.  The Dallas Stars are in the mix.  High offensive production is great, but if it’s just chaotic high-event play at both ends of the ice, it might not be as effective at returning positive goal differential. Not unless your goalie is always the best player on the ice.

The Leafs are sixth worst in Corsi Against per 60 minutes. Most of the teams in this range are very bad; however, Pittsburgh, Dallas, Montréal and Nashville are in the mix too.  Shot rates fluctuate over only a few games, and the effects of a tough or easy schedule, one or two players’ injuries, etc. can have a huge effect.  In the Leafs case they are breaking in a lot of new players.

If you look at the Leafs games individually, hoping for a trend of tightening up in the shots against, you will be very disappointed. Most of the better games by that measure were early ones.  And all we should take away from that is that they aren’t performing at all aspects of the game equally well.

Digging deeper, and looking at just Scoring Chances, the Leafs are fourth in percentage, while in High-Danger Chances, they are second. Split that into for and against and you have the Leafs better on offence than defence, just like in Corsi.

When you look at Goals For, you see something different.  In terms of GF%, they are 27th, and they are in that position because they are much worse at goals against than they are shots or scoring chances against. They have the worst Goals Against per 60 in the league.

About Those Goals For

(Five-on-five, unadjusted, from Puckalytics)

Tyler Bozak is on a hot streak. Every player gets them, and they get cold streaks too.  Shooting is not consistent, no matter how many times a coach laments his players just have to contribute more.  His will fall, and others will rise.

As far as the team goes as a whole, their shooting percentage is 16th, which is the definition of average, and we should not count on seeing various regressions helping out there.  They have the eighth highest goals for per 60 in the league.  There is only so high you can fly.

Conclusion

Go back to the beginning of this post, and I talked about how goals drive wins.  Forget the shots for a second and look at just the goals for and against.

We are taken right back to where the standings page on NHL.com brought us.  A negative goal differential in all situations.  And we all know the major cause, particularly early on: bad goaltending.

But goal differential within a sample of games is still driven by shot differential as well as skater skill and luck.  The system is not innocent.  It isn’t just Freddie Andersen in the parlour with a lead pipe.  It is also the system with the rope as well as the defensive execution with the candlestick.  (I really hope you’ve at least seen the Clue movie right now.)

Judging what you have to do to the team in order to change that high goals against as the Leafs move forward is Babcock’s problem.  But even if the Leafs lay down ten games mostly like the first 13 in terms of Corsi percentage, they are playing a very high risk brand of hockey.

Fulemin said this a few days ago:

This year it’s even more pronounced; the Leafs are the #1 team in the NHL in score-and-venue-adjusted CF60—by a lot—and they’re #23 in CA60. You might assume this was due to the rambunctious young forwards, except their ranks in the same numbers last year were #4 and #23, respectively. I think at least some—by no means all, but some—of our offensive skew is a stylistic choice.

I agree with that.  It is interesting to note that the most successful quarter of hockey last year coincided with the highest overall shot rate for and against.  If you look up at last year’s rolling points total, you can see that that is the second quarter, just before the wheels fell off the goaltending.

Perhaps we have now seen, after the World Cup of Hockey where Babcock’s team played a similar yet more controlled style, the last of the “the Leafs play just like Detroit” comments. I think what we are seeing is a stylistic choice to maximize the highest quality assets, and minimize the lowest. And while Babcock did some similar things to mitigate a weak defence in Detroit, he did not have the Leafs level of offence to do it with.  In his last year in Detroit, the team had the league’s lowest shots against rate.  He was covering over offensive weakness as well as defensive with one of the tightest systems in hockey.

What he has created in Toronto is something superficially similar to the Dallas Stars. While Dallas has less overall skill on offence, they play at similar rates.  I think the Leafs have a different offensive style that is less reliant on rush chances, and the biggest difference in their season stats this year is that the Stars are very poor in Scoring Chance percentage.

In the future, we will find out how well Babcock can balance defensive weakness and goaltending that isn’t Price-like with an intensely fast and skilled offence. You can’t just call up the St. Louis Blues and have them sub in for you when you play a tough offensive team like the Kings.  You have to figure out how to do it yourself.

Throughout the season last year, the shots against declined.  If they can do that this year, they might have a good probability of seeing success at the end of the season.  Or the wheels might fall off.  That’s how it works.