This year’s Leafs offense is historically improved

    The 2016-2017 season is more than a quarter over and the Toronto Maple Leafs are second in the NHL in goals scored. The additions of Auston Matthews, William Nylander, Mitch Marner, and a healthy James van Riemsdyk have a lot to do with that. Last year’s Leafs, however, were not as dynamic. Third-last in the league in goals scored, in fact. A variety of explanations can be offered, but those explanations are not the focus of this post.

    Let’s instead see if the data matches our instincts: Has Toronto’s goal-scoring ability actually improved that much since last year? Have other teams improved similarly? To answer these questions, I chose to use expected goals scored per 60 minutes (xGF60) rather than actual goals scored. If you’re comfortable with that statistic, feel free to skip the next two sections.

    How do expected goals work?

    Using NHL-provided game data like shot distance, angle, and type, whether it is a rush shot, a shot off of a rebound, or a shot from a player’s offwing, expected goals calculates … well … the expected number of goals. Borrowing a quote from this excellent piece from loserpoints (a.k.a. Alan) on NHL Numbers:

    [Expected goals] calculate the likelihood that each shot taken will lead to a goal and uses that to calculate the expected goals for a team. So, if a team generates 20 shots that each have a 5% chance of being a goal, they would accumulate one expected goal.

    @DTMAboutHeart‘s expected goals model, whose data I use for this piece, also takes shooter talent into consideration. You can read more about how shooter talent is calculated here.

    Why use expected goals instead of actual goals?

    Goal scoring involves luck. Example: the New York Rangers have scored 61 even strength goals on 526 shots this season, which is good for a shooting percentage of 11.6%. Only one team has shot higher than 10% over a full 82-game season since 2007-2008. Most teams finish the season between 6% and 9%. Even if we assume they are getting more shots from dangerous locations and have above average shooting talent, they will very likely shoot below 10% come season’s end. Over 24 games, that’s at least nine fewer goals, which is 0.375 fewer goals per game.

    Expected goals remove the noise in situations like the one above so we can compare each team’s true goal-scoring abilities. It works well for a quarter season’s worth of games, and especially well for full seasons. In fancier terms, @as per DTMAboutHeart:

    At the team level, xG has the same predictive power at the 20-game mark as score-adjusted Corsi (CF%) and Goals For (GF%) but proves to be a far more superior predictor of future goals past that mark.

    xGF60: Last Year vs. This Year

    I created the following slopegraphs using DTMAboutHeart’s even strength expected goals data (via herehere, and another file exchanged via email) and‘s time-on-ice data. Slopegraphs, popularized by data viz pioneer Edward Tufte, are especially effective at comparing groups (NHL teams) between two points in time (last season vs. this season).

    Here are the slopegraphs for 2015-2016 vs. 2016-2017. The left half of the figure contains all teams with improved offenses this year, hence the increasing slope. The right half contains teams with fewer expected goals scored per 60 minutes this year, hence the decreasing slope.



    Whoa. Toronto, represented by the blue line, stands out immediately as the most improved offense at even strength. It isn’t as obvious via the slopes, but Dallas’ offense (represented by the green line) has suffered the most this year relative to last year. You’ll notice these teams at both ends of the bar chart below, which displays the changes in expected goals scored that was previously encoded in the slopes of the slopegraphs.


    A few quick thoughts:

    •  The Leafs offensive improvement dwarfs everyone else’s. They are expected to score 0.65 more goals per 60 minutes of 5v5 ice time. That’s more than double the next most improved team!

    •  The Edmonton Oilers offense, led by a healthy Connor McDavid, has improved a ton as well.

    •  The Rangers, who lead the league in actual goals scored per game, don’t come close to Toronto’s xGF60 improvement, despite similarly proficient offenses in 2015-2016. We might expect some regression.

    •  Dallas’ offensive drop-off is stark, but it is likely injury-aided and isn’t separated from the pack like Toronto’s ascent.

    Toronto’s offensive surge this year is absurd considering where they were last year. Let’s see how it stacks up historically.

    xGF60: Historical Year-to-Year Changes

    Here are the same slopegraphs, except for every season since 2007-2008, which is when expected goal data becomes available. Only the teams with the largest yearly increases or decreases in expected goals per 60 minutes are shown.


    And here is the corresponding bar chart:


    This year’s Toronto Maple Leafs are the single most improved team in terms of 5v5 offense in the expected goals era. They are first out of 300 teams (ten seasons x 30 teams) and by a sizeable margin, too. The second most improved team is the 2014-2015 Minnesota Wild and they are 0.17 expected goals per 60 minutes behind.

    A few quick thoughts on the other teams:

    •  This year’s Oilers are the 11th most improved team. Connor McDavid is decent.

    •  Last year’s New York Islanders are at the other end of the spectrum with the largest decrease in expected goals scored per 60 minutes. This surprised me.

    •  The difference between Ron Wilson’s last two full Leafs seasons shows up on the wrong side of the leaderboard at second worst. Who would have thought a team dedicating major minutes to this group (plus Joey Crabb, Tim Brent, Fredrik Sjostrom, Keith Aulie, and others) wouldn’t score many goals?leafs20102011

    Since we’re isolating the 26 teams with the largest yearly swings in expected goals scored, every team is bound to have a good storyline. Too many good storylines for an article that is already approaching 1,000 words. I encourage you to dive into those charts and share your findings in the comments or elsewhere.


    Will this year’s Leafs end up the most improved offensive team ever? We can’t be sure. We only have expected goals data since 2007-2008. Toronto’s rookie-laden team might fade as the grueling 82-game schedule takes its toll. Lou Lamoriello & co. might decide to sell prior to February 28th’s trade deadline. They are, however, head and shoulders above the other 299 teams in our dataset. They have room to fall and still retain their current title.

    Whatever happens, this year’s Toronto Maple Leafs will remain one of the most exciting teams in the NHL, which is something that cannot be said for last year’s edition.