How good do you feel because your team has a new coach? I mean, really…it’s almost like a new-car smell. So many possibilities – This time, things will be different. With the exception of coaching changes due to disastrous, unexpected things, the typical hockey fan was ready for that moment, and were happy to see the coach go. But is that eagerness for change based on real results?
Ever since the Winnipeg Jets finally sent Claude Noel packing, I’ve become interested in how shifting the guy behind the bench might manifest itself in the team in front of them. I mean, if I really wanted to save myself some time, I could just ask Neil Greenberg how he feels about Dale Hunter (as we’ll see in a later post, that feeling is totally justified). Instead, I took the apparently audacious move of analyzing Paul Maurice before he had even coached a game, and suggested Maurice might be a good guy to bring in because Noel clearly was not assessing players particularly well. That will be an important caveat to this entire exercise, in fact: it’s better to replace an obviously incompetent coach with at least a mediocre alternative.
In addition to Noel’s struggles, I also found a bit of data that suggested that Maurice’s squad did a bit better than his replacements, or who was replacing, in the past. I compiled that data by looking at all the instances in which he coached at least 20 games as did his replacement within the same season. As you can imagine, once I did this I became intrigued as to whether this kind of slight uptick was common to coach replacements, or an indicator of Maurice’s influence.
As I’ve hinted to others on social media, I’m sitting on top of a large amount of historical data, scraped from over 40,000 Hockey Summary Project pages, that I still can’t find a good home for. It also provides a massive amount of potential for research, particularly in the way I compiled it, which includes game-to-game coaching data in addition to period-by-period shot data, all back to 1952-53. For the purposes of this research, I wanted to look at all coaching changes where the coach and the replacement coach logged at least 20 games within the same season. I came up with 141 instances of this occurring, spanning across those 60+ seasons, and even at a rudimentary glance the data was telling.
|2pS% Diff||C1 PDO||C2 PDO|
“2pS% Diff,” or as people more familiar with statistics would refer to as 2pS% “Delta” (little triangle symbol, meaning “change”), is the difference in my historical possession measure using shot totals from the first two periods (in order to avoid score effects). A shift in possession of 0.6% is like moving from 50% possession to 49.4% possession…with a sample like this, that’s a pretty good signal, albeit it a minor one. While the NHL can range in team possession from 40-60% (per the Rule of 60-40), most seasons tend to fall in the 45-55% range – which makes 0.6% a potentially important bump.
Now, there can be any number of reasons for why we’re seeing a positive bump. It could be tactics, it could be player usage, it could be positivity or respect in the locker room, it could be fewer/more bag skates. It could be all of these. What’s pretty clear, though, is that the positive feeling you get about dropping that old coach is not unfounded.
On the other hand, the PDO above is worth mentioning as well. PDO is the addition of team shooting percentage and team save percentage (usually at even-strength, but I’m not able to do that across the bulk of my sample…it works basically the same anyway). It’s usually multiplied by 10 to make it a 3- or 4-digit number. Though in extreme cases it can suggest team talent of up to 1% above/below 1000 (or 990-1100), the league average pulls heavily to 1000. This means that most teams above 1000 will likely regress to 1000, and those below will regress up to it. What it suggests above is that the dearly departed coaches might not only suffer from inferiority from their replacements, but also some bad luck as well. a 984 PDO is quite low for any team; with such a wide variety of teams and coaches as this data provides, that’s a strong indicator that a string of bad luck tends to get one fired. The replacement coach, then, not only gets the boost of replacement, but also regression.
There’s a lot more to be done with coaching change data that you’ll see out of me the next few weeks, but even this preliminary look says a lot: switching coaches can help.