• Poulin: “Analytics talk doesn’t make a lot of sense”

    by  • March 31, 2014 • Hockey • 26 Comments

    The Maple Leafs front office is basically a parody of a parody at this point. We’ve all been enjoying the hubris throughout the season. My view was basically that their seemingly crazy views were going to catch up with them eventually, whether it’s this year or next year and that ultimately, the people who bought into the numbers were going to be vindicated by a Leafs team that never accomplished anything of note before having a round of firings and trying again. With the eight game losing streak, that seems like it’s going to happen now.

    Dave Poulin went on TSN 1050 today and was asked about whether maybe, just maybe, there might be something to this analytics claptrap. Response? “Our shot differential over the last 8 games is better. So the analytics talk doesn’t make a lot of sense.” Hoo boy.

    You see, one of the fundamental tenets of analytics is that teams play differently when they’re leading or trailing a game and that it shows up in their shot counts. This is not a new idea. It has been around for years. If the Maple Leafs are publicly rejecting the utility of analytics (and they have, repeatedly), you would expect that they would be familiar with the concepts. They don’t seem to be.

    For the season as a whole, the Maple Leafs have a 48.8% Corsi% when they’re trailing, a 41.3% Corsi% when they’re tied and a 38.7% Corsi% when they are leading. This graph contrasts how much 5v5 time the Leafs spent leading/tied/trailing through their first 68 games with how much time they’ve spent leading/tied/trailing through their last eight games.

    One of the most valuable sports franchises in the world. Managed by people who are apparently completely unaware of how the most completely straightforward metrics about their business work.

    Email Tyler Dellow at tyler@mc79hockey.com

    About

    26 Responses to Poulin: “Analytics talk doesn’t make a lot of sense”

    1. Ian
      March 31, 2014 at

      “One of the most valuable sports franchises in the world. Managed by people who are apparently completely unaware of how the most completely straightforward metrics about their business work.”

      Astute observation, Tyler. Although in a way that’s to be expected. There’s simply never been market discipline in this town, and as a result, the team has been allowed to flounder with relative impunity. I’m not suggesting anyone likes losing. But the imperative to win is a lot clearer when teams start losing money as a result.

      • Derek
        March 31, 2014 at

        @Ian

        “But the imperative to win is a lot clearer when teams start losing money as a result.”

        That’s an understatement. The imperative to win is a lot clearer when your existence may be at stake. The Leafs would need to be unimaginably bad for a horrendously long time with series of comically awful management teams before relocation of the Leafs would be a possibility. LA Kings or Minnesota Wild? They’re not nearly as safe.

        • Phil
          March 31, 2014 at

          LA Kings? You mean the team that’s been in the same city for 47 years and had awful management for 40 of them? I don’t think they’re the best example.

          • Derek
            April 1, 2014 at

            Awful management, but they’ve still been successful at times. You’re right though, not the best example. Poor choice on my part.

      • TMS71
        April 2, 2014 at

        Reminds me of the chapter about the Cubs in the Scorecasting. They don’t need to be good to draw. Their attendance is the least sensitive to performance of any team in MLB. Leafs aren’t really punished for losing so they won’t try as hard to win. Sure they try, as do the Cubs, but they don’t have the desparation.

    2. PopsTwitTar
      March 31, 2014 at

      Im not that analytical at all…but it doesnt take someone with anything more than basic math skills and a a few weeks of watching hockey games to realize teams with the lead “sit back” and teams that are trailing “make a push”. You dont need to be skilled at Excel to make a pretty simple chart showing the amount of shots for and against when teams are leading/trailing either.

      • Kris
        March 31, 2014 at

        Yeah that’s the thing, the “analytics” of hockey are pretty intuitive at the most basic level (outshooting teams 5 on 5 close score is good) and it’s not like some super scary concept.

    3. Ryan
      March 31, 2014 at

      Is it possible that the front office has a working understanding and appreciation of analytics, but just don’t want to admit it because it shines a light on their glaring inability to build and/or coach a team?

      I just blew my own damn mind.

      • Ian
        March 31, 2014 at

        I might have thought they knew what they were talking about. But then I attended a conference at Rotman a few weeks ago where Reid Mitchell was on a panel with (among others) Alex Rucker, who does analytics for the Raptors and is a very smart dude. Unless Mitchell’s out of touch with what his own scouting staff is doing or the rest of the team outside of the scouting organization is really on top of this stuff, it was pretty clear there was not a scintilla of curiosity, let alone understanding, about this stuff.

    4. googligoo
      March 31, 2014 at

      Analytics isn’t the Leafs problem.

      As has been the case for decades, it’s not enough talent that is the problem. Their core is a joke compared to the league’s top teams.

      • Marcus
        April 2, 2014 at

        Is St. Louis’ talent *that* much superior to ours? Backes, Steen, Oshie, Pietrangelo, Shattenkirk, J-Bo, a healthy Tarasenko, Jaden Schwartz et al., Halak/Elliott in goal had them among the top 2-3 teams in the league all season long. Is that *way* more talent than Kessel, JVR, Lupul, Kadri, Phaneuf, Gardiner, Rielly, Franson, Kulemin, Bozak, Bernier/Reimer?

    5. michaeld
      March 31, 2014 at

      If the Leaf’s positive shot differential while losing is explained away by score effects (i.e., they are outshooting now only because they are losing), can you just quickly show that their negative shot differentials for the first 68 games wasn’t the result of score-effects benefiting their opponents? (I.e., that they weren’t being outshot just because they were winning). I suppose the graph shows that their corsi % while tied was pretty bad all year… is that enough? The reason I ask is because disproving Poulin’s dismissal of analytics means not only explaining away his “evidence” (i.e., the inverse relationship between shot differential and winning) but doing so in a way that doesn’t rebound to undermine the argument that they were terrible even while they were winning.

    6. jeffgm
      March 31, 2014 at

      We know that corsi hold across the league and team but has anyone tested how corsi holds for the leafs this season? To answer this, I ran a logistic regression on corsi close over this season to see if over the 70 or so games how well corsi close predicted the leafs win/loss record. For the leafs, what I found is that corsi close failed at significantly predicting when the leafs win or loss any particular game. It probably works for most other NHL teams as we know corsi vs winning holds across the league but for an individual team like the leafs, this correlation fails for this season. For you stats heads, the Chi Square was 0.596 and was greater then the p-value of 0.440 meaning that random chance better describes the leafs record then Corsi Close.

      This suggests, that perhaps luck driven SV% and SH% and special teams etc drove the leafs win/loss success and not corsi close. But anyone claiming Corsi close was some predicting this is overstating the statistical significance.

      • dawgbone
        April 1, 2014 at

        Corsi doesn’t tell you who won or lost the game.

        What it does do is give you an idea of how a team will perform going forward. Sv% and Sh% is, for the most part, random. Teams don’t consistently get .920 goaltending and shoot 9% game in and game out.

        When you have to rely on sh% and sv% to win games (Which the Leafs have basically done all year), you are going to find points where one or both drop. When you can’t generate the shots to make up for those drops, you are going to have big time issues.

      • Kris
        April 1, 2014 at

        It’s generally better at predicting future outcomes or say the second half of the season than explaining past outcomes. So if you were to predict a second half slide for the Leafs based on their team fenwick/corsi in the 1st half you’d likely do better than if you were to predict results similar to what their current W/L record was at the time. Fenwick is definitely more predictive than their goal differential or W/L record at the halfway point.

      • Kris
        April 1, 2014 at

        And yeah, obviously dawgbone is right that sh% and sv% and goal differential are better at explaining the past results, even if those stats are not the best at predicting the future since they tend to have significant variance.

    7. jeffgm
      April 1, 2014 at

      Re-read Poulin’s quote because you guys are going into a different direction now.

      Further, Fenwick Close offers a 34% correlation with winning versus shot differential (with score effects) that has a 27% correlation to winning. The score effects give a 4% increase in an already soft correlation. In terms of predictive reliability (r) that improvement of close is even less.

      • jeffgm
        April 1, 2014 at

        That should be a 7% increase and not 4. And in terms of predicitive reliability fenwick close is .337 vs shots which is .326. That is a difference of 1%.

    8. Scrivy
      April 1, 2014 at

      It’s funny how in hockey, you have to “have played the game” or your analyses are dismissed.
      Generally playing the game doesnt correlate with a knowledge of statistics.
      It’s easier for guys who played the game but don’t know statistics to dismiss stats out of hand, rather than admit that they are worse at analyzing the game than some math nerds.

      It’s the hockey culture. Edmonton and Toronto, it appears.

    9. beingbobbyorr
      April 1, 2014 at

      Why does anyone put any faith in GM’s public statements about MoneyPuck?

      I assume that the truthful answer — were they motivated to give it (and they’re not) — would be “we’re studying the hell out of this stuff with our own versions of Bill James, Paul DePodesta, Tyler Dellow, etc., and — whatever (if anything) we’ve found so far — it’s not in our interest to disclose a damn thing (yet)”.

      • Martin A.
        April 3, 2014 at

        Their ignorance of stats is shown in the decisions that they have made and continue to make. Keeping Bozak and letting Grabo go. Believing Clarkson was worth 6 MM/yr. Blaming Reimer. Etc. etc.

        • beingbobbyorr
          April 3, 2014 at

          That doesn’t prove “ignorance of stats”. Bad hockey decisions occurred way before Microsoft Excel was invented.

          • tony
            April 5, 2014 at

            That’s the point. “Before Excel was invented” people could get away with making those bad decisions because the metics we have now didn’t exist. Having this proven metrics right in front of their eyes, dismissing them completely, and fielding a horrible team which analytics could have prevented is inexcusable.

            • beingbobbyorr
              April 5, 2014 at

              Right now these metrics are merely interesting. They’re light-years from being considered “proven”.

    10. Pingback: Toronto Maple Leafs Fire Claude Loiselle & Dave Poulin, Hire Kyle Dubas As Assistant GM - Editor In Leaf - A Toronto Maple Leafs Fan Site - News, Blogs, Opinion and More

    Leave a Reply

    Your email address will not be published. Required fields are marked *