The Toronto Maple Leafs continue to be an anomaly and amaze us all

2012111-maple-leaf-license

During the third period of last night’s Leafs game against the Kings, the Leafs held a 3-2 lead and had a faceoff in the defending end of the rink, to the left of James Reimer. The Kings had Anze Kopitar on the ice, an elite forward and faceoff-man. The Leafs had Tyler Bozak and Jay McClement.

A discussion took place between Chris Cuthbert and Ray Ferraro of TSN, discussing the merits of each centreman taking the draw in that situation: McClement is a better faceoff-man and had had an excellent night against Kopitar on the night, going 6-2 on draws. Bozak, on the other hand, is right handed, which makes it easier to tie up the opponents stick on the forehand when all you want to do is kill some time on the clock.

To me, the decision you make there is steeped in analytics.

Brian Burke talked during the Sloan Sports MIT Conference, predictably, about faceoff issues, and problems that arise when the NHL records the data. The NHL records a faceoff as a win or a loss for either centreman with no worry for the situation, whether his wingers helped him win the draw, or whether the player was tied up. Having a better idea of which faceoff men are better in which situation is important, to a degree, if it can help you go from a 50% faceoff team to a 52% faceoff team. Those marginal possessions during a game can add up to one or two goals over the course of a season, and the Leafs are in such a competitive conference that finding an edge in every aspect of play can help.

HURRICANE EPSILON

Unfortunately, analytics has taken a little bit of a hit over the last 18 months, and mostly due to the success of the Toronto Maple Leafs. This presents a very interesting conundrum not just for Leafs fans, but for Leafs management. As I’ve written before, there’s no worry for Dave Nonis to worry about what a Corsi number is until a Corsi number can tell him something about his team (despite the 2008 Vancouver Canucks suffering a precipitous fall similar to what most analysts like myself were predicting for the Maple Leafs this season).

We are also past a point of the season where percentages and shot differentials can tell us too much. Playoff races and playoff series are exercises in small samples. People weren’t lining up to pick the Maple Leafs to have a chance against Boston a year ago, and the Leafs came oh-so-close to one of the most stunning playoff upsets of the day. There are 14 games left in the season and I’ve resigned to lose my bet to Twitter dot com’s @hip2jive that I made during the preseason. What follows now is luck. Good teams usually win in the postseason, but there are usually six or seven good teams that have a chance. Last year I went all-in on Boston, Chicago and Los Angeles in my playoff pool and won big. I’m having a difficult time picking against either team this time around, and it certainly helps if there are no big upsets along the way, but those happen too.

Because… there’s still nothing I see in this Maple Leafs team that makes me believe they’re on the same competitive scale as the top teams in the conference or the league. Taking two wins in three games in four nights on that California trip isn’t something that many teams have done this season. Tampa Bay got slaughtered during their trip. Boston won just one of their games, as did Columbus. Montreal won a single game, in the shootout.

And yet, I watched that game against Los Angeles and again saw the same team I’ve watched for the last 18 months—they were hemmed in their own zone, and won thanks to 3 or 4 big saves from James Reimer. Blue & White Disease doesn’t yet have the scoring chance numbers up on their site from the game against the Kings, but in the 3-1 win against Anaheim the Leafs were out-chanced 28-12 and by San Jose 34-13, which gives you some indication of the troubles Toronto has faced this year. At 5-on-5, the Leafs have been out-scored by five goals this season. They have an overall goal differential (shootout-excluded) of minus-9. They’ve won a higher-than-expected number of close games and have relied on situational scoring and goaltending. They win games by an average of 1.7 goals and lose by an average of 2.2. They’ve picked up four more points than the average team in the shootout. Even if you aren’t looking at the Corsis or the PDOs, you’re looking at a team that is both hanging onto its position in the standings and somehow also making a charge. Tampa is struggling for wins since the Olympic break, and the Leafs are three points up on them.

The part of me that is entertained by the chaos is as frustrated as the weather researchers tracking Tropical Storm Epsilon back in the 2005 Atlantic hurricane season. Via XKCD:

epsilon_and_zeta

(This is just a snippit. Go to XKCD for the full comic. It’s an excellent online strip)

There’s a similarity here: conditions aren’t favourable for storms in the winter, and yet the storms persisted. There wasn’t much left to write about. That’s how it’s become with the Leafs. The Leafs approach their decisions so illogically compared to a team like Chicago or San Jose in regards to deployment and tactics, and yet are having similar success within their conference. You can’t turn around and say “oh, Jay McClement needs to carry the puck in more because X number shows that the Leafs are getting outplayed when McClement’s on the ice” because there’s a winning formula here the Leafs have somehow stumbled upon that can’t exactly be messed with this late in the season.

The team looks drastically different than it did in the fall, I would add. James Mirtle wrote up a post about how Randy Carlyle isn’t using his tomato cans as much, and got rewarded with an overtime goal from Troy Bodie, who isn’t a player with a whole lot of above replacement talent, but is the kind of player you want in the lineup: quick, with some skill, and potential for three or four goals throughout the year in the right situation. Jake Gardiner and Morgan Rielly both have more freedom in the offensive zone.

If I’m looking forward to the future, I have to beat the same old drum. The Leafs have a disproportionate number of wins in one-goal games. They have the second-worst possession statistics in the league, ahead of only Buffalo. Jonathan Bernier is on pace to have the best season ever for a Maple Leafs goaltender, in his first season as a starter at any level since 2009.

So… despite disfavourable conditions, Hurricane Epsilon continues to roll. I can’t help but predict misery ahead, but they’ve seemed to do well without me. Not to say that forecasting hockey is as accurate as forecasting weather.

And yet…

December 5, 2005 at 10:00 pm EST:

We have said this before at times during the past several nights… only to have Epsilon make a comeback the following morning… but Epsilon really does not appear as strong this evening as it did this afternoon.

December 6, 2005 at 10:00 am EST:

Epsilon appears to still be a hurricane… but just barely.

December 6 2005 at 10:00 pm EST:

The end is in sight. It really, really is. But in the meantime, Epislon continues to maintain hurricane status.

And so on and so forth.

Stats via ExtraSkater and NHL.com

  • I’ve often wondered about what I am going to post re: the leafs.

    Normative data falls into that 95% range (roughly the mean +/- 2 SD’s), is it possible that the Leafs are an outlier and can be a Team that can sustain this type of play?

  • v4ance

    This may be part of my DNA now as a lifelong Leaf fan but, every time the Leafs win, I keep waiting for reality to return in the next game.
    The current edition of the blue-and-white does little to discourage my pessimism but I have to say that the win in LA was a terrific surprise.
    Sunday in Washington? I’m ready for just about anything.
    What a team!

  • “They’ve won a higher-than-expected number of close games and have relied on situational scoring and goaltending.”

    Bingo! You almost have it there.

    The Leafs are a very talented, but young, sometimes lazy and undisciplined team.

    They have good goaltending, speed and good enough scorers that they can actually get goals when they need them to win close games.

    Rather than call this chaos, why not refine your modeling and analysis to account for and explain what the Leafs really are?

    Lets look at the elite in team shooting percentages:

    1 St. Louis Blues 10.93%
    2 Colorado Avalanche 10.28%
    3 Pittsburgh Penguins 10.27%
    4 Toronto Maple Leafs 10.21%
    5 Anaheim Ducks 10.19%

    Lets now look at the elite in team save percentage:

    1 Boston Bruins 0.925
    2 Los Angeles Kings 0.922
    3 Colorado Avalanche 0.919
    4 Montreal Canadiens 0.918
    5 Toronto Maple Leafs 0.918

    Resist the temptation to turn that into PDO (team save percentage + team shooting percentage) and say the Leafs are “lucky.” (PDO is flawed, Pittsburgh has a had above average “luck” since the 1990’s. This could not happen if there was true randomness)

    Good shooting and goaltending…it’s how the Leafs win.

    “Because… there’s still nothing I see in this Maple Leafs team that makes me believe they’re on the same competitive scale as the top teams in the conference or the league.”

    The Leafs are talented enough that they can beat many top teams. They have trouble with teams that can play good defense: In the East that’s Boston. Columbus also gives them trouble. In the West, the Leafs can’t deal with St. Louis or San Jose.

    But if the Leafs met Pitsburgh in the second round, the Leafs might give the Pens a good scare or even beat the Pens. Over the last two years, the Leafs actually won a good number of games against the Penguins.

  • Regarding the episilon analogy – at some point scientists have to consider that their forecasting model and theoretical framework is incomplete or perhaps misleading. Einstein, Newton and countless other scientists improved the existing framework when “reality” did not match the model. [Gravity worked well as defined by Newton and Einstein evolved it so that the new view was consistent with Newton’s view].

    I find some in analytics are holding onto what maybe an incomplete or misleading model. And excuse the limitations of their model by saying that such and such a team just got lucky and will regress because that is what my model says.
    [There are statistical “tests” to determine how much chance or skill is at play at 60 or 70 or any number of games into the season and relying on luck to justify the existing shot differential is questionable at this point. Tom Tango has done pioneering work on luck vs skill].

    At this point in the season, at minimum there needs to be a healthy skepticism that luck is sole factor because the probability that luck is influencing outcomes is decreasing each game and skill effects are now beginning to dominate the game outcomes.