Why the Leafs are 6-1

This post from Phil Birnbaum yesterday is well worth a read. It’s interesting and while I disagree, I think it’s important that data-driven people look for reasons why the Toronto Maple Leafs won so many games last season despite being out-shot.

Advertisement - Continue Reading Below

At 6-1 to start the season, it’s going to take more than theories, charts, and graphs to show people that what the Maple Leafs accomplished last season was unsustainable. I will point out that the things driving the Leafs to victories last season, chiefly a high 5-on-5 shooting percentage rate, aren’t the same things driving their success this season.

It turns out that seven games is far to little to judge a hockey team’s successes and failures, but let’s delve into the Birnbaum argument anyway.

For those new to the discussion—congratulations are in order for finally plugging into the Internet—there’s a great post at Grantland from Sean McIndoe called “Collision Corsi” that explains the two ‘sides’ to the great debate surrounding our prognostication for the 2014 Toronto Maple Leafs.

Birnbaum presented the idea that because the Leafs sat back in a defensive shell, and had a high shooting percentage, they take on the identity of teams that are up by two goals in a game. He’s not a hockey watcher, but he does point out that, between 2007 to 2013, teams appear to shoot the puck much better when up 2 goals than down 2 goals:

Advertisement - Continue Reading Below

7.60% … down 2+ goals
7.75% … down 1 goal
7.52% … tied
8.40% … up 1 goal
9.19% … up 2+ goals

he concludes:

regardless of the details of the explanation: how you play does indeed seem to influence shot quality.  It’s not all just random.

Which is true. “Score effects” is a principle reason why analysts like to look at “score close” or “score tied” metrics to judge teams. The average team, as calculated by Birnbaum, will have a 50% Corsi rate in score tied situations (of course), but 45.1% when up by 2 goals and 57.0% when down by 2. This does carry over to goal-scoring as well.

The “defensive shell” theory may mean that if you’re hemmed in your own end, you have a better chance of capitalizing on a mistake and scoring off a rush. Something like this goal by the Sabres against the Leafs last season, that came at the conclusion against a long Toronto shift in the Buffalo end:

I think there are some good reasons why teams shoot better when they’re up by 2 goals, but it has nothing to do with the strategy employed by the leading team and more to do with the trailing team. The team that’s behind is going to play more aggressively, sacrificing high percentage plays for low percentage plays, tipping the scale in favour of more risk for more reward. Tthe graphs in this great post by Driving Play on score effects show just how dramatic the incentives are for each team later in the game. Preventing a goal when down by a goal at the 55-minute mark is worth about 0.3 points in the standings, but scoring a goal would be worth 1.0.

(Since Hockey Analysis does not include 6-on-5 situations, you can throw out the idea that teams that are leading have higher shooting percentages because of empty nets)

Advertisement - Continue Reading Below

Would the Leafs’ supposed strategy be a good one? Probably not. Birnbaum qualified that by pointing out the goals for and against per 60 minutes by teams in those situations:

2.42 – 2.32 … down 2+ goals
2.39 – 2.26 … down 1 goal
2.21 – 2.21 … tied
2.26 – 2.39 … up 1 goal
2.31 – 2.42 … up 2+ goals

“being down actually gives you a small advantage in future goals” he writes. I think that there have been enough goals scored between 2007 and 2013 for us to conclude that there is a very real advantage to the team that is down by a goal. Based on Pythagorean expectation, the average team goes from a 39-win team while leading to a 43-win team while trailing. In Moneyball, writer Michael Lewis invokes the “balance of strategies” that tip the scales, whether strategic or psychological, into the favour of the trailing team.

Ultimately, the arguments made by Birnbaum can be refuted, and have, by Eric T. at Outnumbered and by draglikepull at Pension Plan Puppets. I’d like to see a few more of these written up, preferably with data, but I’m still not convinced that the Leafs style of play is sustainable. It’s possible that the Leafs have stumbled onto some method of shot quality not seen since the dawn of the salary cap era, but it’s unlikely.

But the Leafs are 6-1, aren’t they?

Yes, they are, and they’re also not winning this season the same way as last season. While the Leafs did win despite being out-shot 38-14 against Minnesota, only one of the goals they scored was 5-on-5. Toronto are just +2 this season at even strength, with much of their goal differential coming by way of special teams. I’m getting the next few figures from Hockey Analysis here.

Cosmetically, the Leafs are third on the penalty kill at 88.9% and third on the powerplay at 33.3%. Powerplay rates appear to be up in the early going than previous years, but that could be a sampling issue. Eventually, the law of large numbers are going to digest the figures of every team and bring them all within a closer range. The Leafs are probably more likely to finish the season with the third best powerplay than the third best penalty kill. I made the point in yesterday’s game preview about the team’s overall save percentage on the PK.

Percentages are a big driver of success in a small sample, and while the Leafs were one of the best teams in the league last season at preventing shots on the penalty kill, they’ve sunk to 25th in the NHL in that regard, propped up by a .953 save percentage on the PK. That’s tops in the league. By extension, the team is fifth in shots generated on the powerplay, ninth in shooting percentage and fifth in goal scoring, so it’s more likely that their current powerplay output is a 

Advertisement - Continue Reading Below

But you get writers and tweets all the time about how the Leafs are “confounding” stats people. Not exactly. We knew going into the 48-game season that we’d run into results inconsistent with observations over 82-game seasons. The message was always going to be to avoid the 48-game sample trap and improve in the right areas. The Leafs look somewhat more competent at evens this season, the game against Minny aside. Carlyle admitted at the start of the season the team needs to be better at puck-possession and he stressed it again after the Minnesota game. I think he’s aware that the team probably cannot continue to win if his team played at evens the way they did in 2013. 

The Leafs, however, have gone from being a team with a PDO that was standard deviations from the mean to sitting at 8th in the league. Those are even strength numbers, but currently, the Leafs’ special teams are driving the bus.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

  • Is it possible that due to great special teams the leafs often find themselves up by a few goals in a game and than play like they are up by a goal or two (because they are up by a goal or two).
    For example last night we played almost 8/20minutes of the 1st period on the PK where the Wild generated a bunch of shots. Despite this our special teams (and a bit of luck) put us up by a goal. I also noticed this in the Nashville game, the leafs weathered a lot of point shots, than capitalized because of luck/PP and than sat back instead of trying to gain access to the zone and possibly give up an odd man rush.
    Basically… the leafs are getting lucky, but it is what they are doing with the luck which is skewing their advanced stats. Kind of like poker, its all luck but the leafs are better at folding hands and picking there bets than other teams.

    • I’m not sure how much I would include luck into the equation, definitely last night wasn’t a great game though. I think you’re right about them playing with a lead though. I believe they played a significant amount with the lead last year and this most likely negatively impacted corsi nums. While ahead, the Leafs try to get it out and cycle it around the offensive zone if they can, without actually taking any attempts on net unless they have a prime chance. Mcclement does this quite a bit.

  • The kids over at PPP were breaking down this birnbaum article as well, and birnbaum was replying to some of the criticisms. He emphasized that he is not positive on some of his hypotheses, but he is pretty sure that the Leafs results last year are most likely not mostly based on luck (as many seem to be saying). It’s just good to see someone putting some alternate explanations out there rather than lazily concluding that since the Leafs had high sh% and poor corsi numbers then they must be just getting lucky.

    • You may disagree with the people at TLN and PPP that write about stats but they are definitely not doing anything lazily. The number of hours spent not just marshalling arguments and information but actually collecting data dwarves the effort put in by most that can’t even be bothered to even begin to try to understand the concepts they are presenting.

      • At least Cam and TLN are reasonable about most of it. Most of you PPP folks are way too obnoxious and unable to handle criticism. You also tend to make weak conclusions that tend to support your collective biases. Look a little bit deeper and you’ll find that it is not just luck when results don’t follow your narrative.

        • Meh, we give what we get. It’s really chicken or the egg about which came first but we’ve been getting hounded about being too negative about the Leafs basically since day one despite having a pretty good record of being right about the Leafs’ decisions.

          Show your work for once. Your lot complains a lot about how you have to dig deeper and then you never do.

  • MaxPower417

    The part of his article that struck me as the most significant was the claim that CF% and SH% over the past 6 years have a negative correlation of 0.22 which isn’t insignificant.

    However upon investigation, that is driven by a crazy shortened 2013 season where they was a negative correlation of 0.47. If you look at the other 5 years, there is almost no correlation in any season and the correlation over the full 5 years is positive 0.06.

    I’m not sure what this means for the rest of his argument, or what the hell was going on last year, but that’s what I found.

    • Sounds like you know enough to make an analysis from a numbers standpoint. It definitely helps to have played the game, especially at a high level to fully understand certain aspects of hockey, but you do make some good points.

    • Sorry if it seemed like I was mis-representing your position. Thanks for stopping by to comment.

      I don’t think that anybody is any more qualified to comment on hockey based on whether they do or don’t play. I’m one of the rare Canadian kids who never played minor hockey, but had the joy of watching my brother wake up for early morning practices and late-night games.

  • Two things to add to this article: check out the opponents’ record to date – the Leafs have a much tougher road ahead; and, secondly, the goal tending so far, touch wood, has been exceptional.
    On another completely different note, I was pleased to see that Jerred Smithson has been offered a Professional Tryout with the Marlies. Smithson is a tough player who certainly can offer more on the skill side than the current Leaf enforcers.
    Looking forward to seeing what happens when the injured and suspended guys return. Interesting times.