Randy Carlyle (Getty Images)
When an NHL team starts the season poorly, the head coach is the first to pay the price. But is firing a team's bench boss a legitimate solution for what ails it? Here's an advanced analytics perspective.
As the fervor dies down from the fever pitch of opening games, NHL teams and their fanbases shift into the time-honored ritual of agonizing over early-season results. A few coaches begin to feel the walls close in, and regardless of testaments of faith by upper management, at least one will be fired in the first few months. Do teams carry out these decisions wisely? What kind of measures can help us determine if it’s a good move? Reaching back to the
THN Analytics stats primer, the best team measures we can use relate to regression and possession. For regression we can use “PDO,” or a team’s shooting percentage plus save percentage (for historical comparisons, I only use the first two periods to avoid the effects of “protecting” leads). It’s often expressed as a whole number like 980 or 1000, rather than their actual values of .980 or 1.000. Teams that are far above or below
a range of about 990 – 1100 pull heavily (or regress) towards that range the remainder of the season. PDO is a great metric for this kind of study because its measure speaks directly to a team’s success in scoring or preventing goals. Possession is currently best measured by Fenwick Close, but we can go further back in NHL history by using a team’s shots-for in the first two periods divided by both teams’ shots-for in those periods, called two-period shot percentage or 2pS%. It
runs side-by-side with Fenwick Close,
has a strong relationship with outscoring, and provides about 50 more years of data. Using these two measures, we can look at a large body of coaching changes in NHL history. Through 140 coaching changes (minimum 20+ games for each coach), the before-and-after of PDO and possession is telling:
Historically, the changes have barely registered an uptick in possession (that 0.4% is worth a little less than one more goal-for), but that PDO shift would be good for about 14 more goals-for. In other words, NHL teams tend to cut bait when bad luck, not necessarily bad leadership, seemed to be the bigger problem. For comparison’s sake, I also put together a complete list of 97 coaching performances where the coaches had significantly low PDOs through the first 20 games but didn’t get canned:
As you can see, whether standing by the coach or replacing them, across a wide variety of situations, teams, and eras, PDO still gravitates toward its average. In terms of possession, it is likely that stronger possession teams elicit better results, and could potentially help their coach’s job security. If nothing else, they give their coach better odds of sticking around long enough for PDO to regress to league average. The volatile relationship between PDO, early-season results, and the potential for firing coaches has caught a number of great leaders in its maw.
1958-59 Billy Reay – Toronto Maple Leafs A legendary name today, in 1958 Billy Reay was a young coach who had just completed his first year behind the bench for the Leafs. Unfortunately for Reay, recently-hired GM Punch Imlach was eager to resume coaching after leading the AHL’s Springfield Indians the previous season. He
gave Reay four games to turn around the Leafs’ slow start. Two ties and two losses later, Reay was fired, and it would be three fantastic years in the minors before he received another chance in the NHL. How did the Leafs fare?
Reay never received the benefit of coaching future stars Frank Mahovlich and Bob Pulford at their peak, and it was Imlach who reaped the rewards of Reay’s effort to convince a long-time minor leaguer named John Bower to tryout with Toronto. By 1969, Imlach had three Stanley Cups and a pink slip, while Reay was leading a Blackhawks squad that would stay competitive through the mid-1970s. Take a guess which coach is in the Hockey Hall of Fame, and which in the Manitoba Hall of Fame.
2000-01 Alain Vigneault – Montreal Canadiens Like Reay, Vigneault was a young coach in his first NHL job, but by late 2000 the Canadiens were languishing in last place. Vigneault might have felt the rug had been pulled out from under him. GM Rejean Houle traded away most of the Habs’ talent during Vigneault’s tenure. To make matters worse, ownership was looking to sell and needed a way to increase confidence in their assets – replacing Vigneault with Michel Therrien became one way to accomplish it.
It took Vigneault a few years to earn another opportunity at the NHL level. His success in Vancouver and New York has since solidified his reputation as a top hockey mind. Montreal removed Therrien by early 2003, only to bring him back in 2012 – the club’s seventh coaching change since 2000.
2006-07 Mike Babcock – Detroit Red Wings Mike Babcock wasn’t fired by Detroit, but I wanted to demonstrate an instance where patience prevailed. In 2006-07, the Wings entered the season with sluggish percentages after a rough first round upset at the hands of the Edmonton Oilers the previous spring. Context is important here: Babcock’s Red Wings went 12-6-2 through their first 20 games, while Reay’s Leafs went 5-12-3 and Vigneault’s Canadiens went 5-13-2. Solid possession work, a hallmark of Babcock’s approach, helps when percentages fail. The possession chart was shifted up 5% on the y-axis to catch how high the Wings’ possession figure was early in 2006-07. The 62% mark is
far above what anyone can expect a team to sustain, but even 55% is top five in the league. Though the 2006-07 Red Wings would not go on to win the Cup that season, the following year, with Babcock at the helm, they would have one of the best seasons in NHL history.
Conclusion Over the course of NHL history, it is not uncommon for coaches with poor win-loss or percentage-reliant results early in the season to lose their jobs. There are certainly coaches that deserve that kind of pressure, but there are ways we can dig deeper to find out if this futility is a trend or if it’s just a matter of patience. PDO can help us see if the percentages are likely to rebound, or regress, towards league-average, and possession measures can give us a sense of whether the coach is making decisions that could lead to success in the future.