The Carlyle Effect Revisited: Randy’s WOWYs

Photo Credit: Brad Rempel/USA TODAY SPORTS

Back in 2014, I wrote an article about Randy Carlyle’s effect on Corsi for players on the Toronto Maple Leafs.  It was the end of Carlyle’s second full season behind the bench after taking over for Ron Wilson part-way, and it seemed like a good time to reflect on how the coaching change had affected possession.  The conclusion was stark: virtually every player who played for both coaches on the Leafs saw their Corsi fall, usually precipitously, after Carlyle replaced Wilson.  The effect was so strong it could be seen mid-season the year both men coached the Leafs for part of the season.

The Leafs have now finished their first full season with Mike Babcock behind the bench, roughly a year and a half after Toronto decided it was time to move on from the Carlyle era.  Since we’ve now got another coaching change to look at, I thought it would be a good time to revisit the idea and see if the Carlyle effect is apparent in reverse.

Advertisement - Continue Reading Below

I’ve collected the 5v5 shot attempt ratio (or Corsi For %) of every player on the Leafs who played at least 20 games under Randy Carlyle in 2014-15 and at least 20 games under Mike Babcock in 2015-16.  That allows for a quick comparison of how CF% has changed over the previous year.  This method isn’t perfect; there are roster and linemate changes between seasons, as well as differences in deployment.  But it gives us a decent overview.


The change in possession between the two coaches is pretty apparent from this chart:


Advertisement - Continue Reading Below

As you can see, other than Joffrey Lupul, every forward who played under Carlyle in 14-15 and Babcock in 15-16 saw jumps in possession, most of them quite large.  You might argue that most of what we’re seeing here is due to roster changes rather than coaching, but there’s a fairly simple way to test that:

What happens when we add in CF% under Peter Horachek, who took over from Carlyle mid-season?


Advertisement - Continue Reading Below

This confirms what we saw in the first chart.  There’s a fairly consistent pattern here of players improving in CF% almost immediately upon Horachek taking over for Carlyle, and then the improvement continues under Babcock.  It certainly looks like Carlyle’s coaching was having a significant negative impact on possession.


We can look at the same numbers for the four defencemen who played at least 20 games under both coaches.


Advertisement - Continue Reading Below

The effect is even more clear here.  Every defenceman who played for both coaches saw large jumps in possession after the coaching change.  Let’s add Horacheck in again:


The result is the same as it was with the forwards: Corsi improves dramatically mid-season once Horachek replaces Carlyle, and then goes up even more when Babcock becomes coach.


I think the charts give a nice visual representation of The Carlyle Effect, but it can be helpful to see the numbers too, so here’s a chart of the difference in Corsi between players who played for both Carlyle and Babcock:

Player Babcock CF% Carlyle CF% DIF
JVR 54.6 41.3 13.3
Bozak 51.6 41.9 9.8
Holland 52.0 42.8 9.2
Gardiner 53.3 44.8 8.5
Polak 49.9 42.3 7.6
Phaneuf 51.3 44.2 7.2
Komarov 53.3 46.5 6.8
Kadri 52.6 49.6 3.0
Rielly 50.0 47.4 2.6
Winnik 49.6 48.3 1.3
Lupul 46.4 46.5 -0.1

The most important thing here is just how big the jumps in Corsi are.  Seven of the eleven players who fit my games played criteria jumped 6.8% or more in possession in just one season.  In fact, the only player who didn’t see a clear increase in their possession under Babcock is Joffrey Lupul, who faced frequent injury problems that made it difficult for him to find his game.

There’s no simple way to average out the effect among the players here because some of their minutes were played together.  But broadly speaking, it’s pretty clear that there’s a very large impact.  In the article that I linked in the opening, I estimated that Carlyle’s negative effect on possession was somewhere in the range of 5-7%.  Over the past couple of years, a lot of people have suggested that I was over-stating the impact, but I think the data I’ve collected here demonstrates that my original estimate was probably pretty fair.

That estimate was based on the difference between Carlyle (a bad coach) and Ron Wilson (a decent coach).  Babcock is better than decent, which is why it’s not too surprising that more than half of the players here saw jumps over 7%.  It’s impossible to positively identify the precise mix of factors leading to the improvements seen here; some of it is due to eliminating Carlyle’s negative effect, some of it is due to adding Babcock’s positive effect, and some is due to lineup changes.  But the overall result is pretty clear and confirms my conclusion from 2014: Randy Carlyle had a significant negative impact on his team from a puck possession standpoint.

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

  • Tommy Cat

    Except that Carlyle won a higher percentage of games he coached.

    The point of playing hockey is winning games not getting a high Corsi. Corsi is such a useless stay anyway. Why focus on that and ignore the fact that the team under Carlyle consistently had higher shooting and save percentages.

    Carlye’s coaching systems were designed to increase shooting percentages at the cost of fewer shots on goal, and increase save percentages at the cost of more shot attempts against. The other coaches listed use systems with the opposite trade off. Either way can work if the team playing it has the right skill.

    Carlyle is not a bad coach. Bad analytics used by people who don’t understand hockey systems makes Carlyle look bad. That’s all.

    • Bob Canuck

      I agree, everyone has an angle and use can use stats to show a negative because you can’t achieve everything with any hockey system otherwise everyone would be doing it.Therefore you pick one thingbase on your teams talent and hope it leads to more wins. If it doesn’t then you try something else. And of course injuries throws all of the stats out the window.

  • DragLikePull

    Carlyle made the playoffs with the Leafs once in a lockout shortened season based on flukey percentages that many people, correctly, predicted would regress the next year.

    His Leafs were disorganised and frequently out of position, constantly giving the puck away to the opposition because they couldn’t manage a breakout.

    • Gary Empey

      Under Babcock the Leafs were organized and rarely out of position, seldom giving the puck away to the opposition. Their Corsi was over 50%.

      Yet still they won only 29 games last year.

      The Leafs problem as Babcock stated before the season started we don’t have enough goal scorers.
      Having possession for over 50% of the time and not being able to score is a disaster.

  • Gary Empey

    Using Corsi all the Leafs improved under Peter Horachek. They also lost almost every game they played.

    Under Babcock there is even a bigger improvement in Corsi. Still we ended up as the worst team in the NHL and able to draft first overall.

    So my question is Corsi a logical, but overrated/useless stat?

  • Gary Empey

    To fellow commentors,

    Corsi is simply a piece of evaluation, like a grading rubric. It doesn’t mean you will pass if that aspect of the team is good but it certainly improves your odds.

    If I remember correctly the Leafs lost 30 some odd games by 1 goal. That’s much improvement from the 9-0 or 8-1 blowouts we saw under Randy and Horachek – being on both sides of the W/L column.

    What Corsi really shows is the team has to puck more than the other team, is moving towards to oppowing net rather than our own and controlling the game. This leads to a greater chance to win.

    Of course we can look at Calgary and see how poor goaltending brings down the team (remember that grading rubric I referenced) or someone like Lindqvist or Price saving a team despite poor Corsi.

    So to answer the question if Corsi is over or underrated, it’s simply a tool to measure probability of success.

    • Stan Smith

      I agree with you 100%. Unfortunately too many “experts” feel it is the only way to judge a team, or worse, a player. One thing this article proves is that being a good Corsi coach does not necessarily translate into being a winning coach, and being a player with good Corsi is more about the team, and the coach, then the actual ability of the player.

  • Gary Empey

    My real question is can Corsi really measure probability of success?

    If a team at the bottom of the league has a good Corsi rating, this stat becomes contradictory.

    Last year after every game, our Corsi stats were given as if they were the defining stat for the game.

    Seriously how can anyone using Corsi ignore the outcome of the games?

    It is very logical to say while we have to puck, that means the other team can’t score.

    It seems to me there are a lot of players in this league that don’t need to possess the puck for very long before a major scoring chance occurs.

    The fault in most Corsi analysis may be using “over the 50% line” as being a sign of good Corsi.

    My guess is using a much higher than 50% would make more sense.

      • Gary Empey

        Nice to see “Rip Van Winkle” has come out of his summer sleep.

        Using Corsi to evaluate the Leafs is similar to judging all women by my first wife.

        She did have a nice Corsi. Some thought she may of been possessed. I thought she made too many turnovers.

        • DragLikePull

          Gary some great lines there. In fact too bad the late Henney Youngman wasn’t around, he would have taken your first wife. For the youngsters in here, watch the great movie directed by Martin Scorcese, Good Fellows as Youngman is the commedian in the scene in which Ray Liotta takes his future bride to see.

    • DragLikePull

      Much higher than 50%, yes and no. This is where CFRel% comes into play. If you’re better than league average that is generally good to great, akin the SV% for goalies.

      My statement is more of an individual output but it does translate to the team(s) as a whole as well. We would have to look at the league to see where everyone else sat in terms of Corsi to determine how much higher than 50%.

  • DragLikePull

    Corsi isn’t a measurement of the probability of success. It’s a measurement of whether teams are doing a good job of possessing the puck at even strength. Teams that do a good job of possessing the puck at even strength generally win more games than teams that do a poor job of it. Teams that get good goaltending also win more often. So do teams with a good powerplay and penalty kill. There are lots of factors that have to come together for a team to be good, but it’s rare for a team to have much long-term success if they can’t control possession.

    It’s not impossible to win a lot of games with a bad Corsi, but it’s unlikely.

    It’s not impossible to lose a lot of games with a good Corsi, but it’s unlikely.

    Lots of things happen over the course of an NHL season but *as a general rule* the teams that win the most are also good possession teams.

    • DragLikePull

      Your opening statement crumbles with your argument. You proved that Corsi is a measurement of probable success, the opposite of your thesis.

      “Teams that do a good job of possessing the puck at even strength generally win more games”

      “It’s not impossible to win a lot of games with a bad Corsi, but it’s unlikely.”

      “It’s not impossible to win a lot of games with a bad Corsi, but it’s unlikely.”

      What you’re saying is exactly what I said yet disputed. It’s a measure of probability to win the game, but not a sure thing.

      To quote Lloyd Christmas – So you’re saying theres a chance!

  • Tommy Cat

    Coaches have to use what they are given. They coach to win in the now. Randy Carlyle got the most out of that group of mutts than any coach could have been able to. Man has a Norris trophy and a Stanley Cup as a head coach …. yet some brilliant minds think he’s a buffoon.

    • FlareKnight

      You know how many trophies and awards Wayne Gretzky has? Didn’t make him a good coach. Just because you know how to do something really well, doesn’t mean you can teach people. So toss Carlyle’s Norris out the window, that is utterly meaningless.

      Carlyle isn’t a good coach in the current NHL. He slipped out of Anaheim and was deservedly fired. He couldn’t get much out of Toronto either. And while the team wasn’t good, management often tried to add guys Carlyle wanted or were “Carlyle guys” to try and make the team better. Didn’t work.

      The man himself was “befuddled” by why the Leafs weren’t playing well. This so called great coach couldn’t comprehend why the Leafs were a bad team the year before he was fired.

      He’s not a good coach and that’s going to show off in spectacular fashion with the Ducks next year.

  • Bob Canuck

    One of the values of data is that some narratives can be assessed using numerical evidence. So far in the comment section of DragLikePull’s post, which I thought was excellent, we have heard some classic narratives: Randy did the best with what he was given; and Randy made the bargain to generate more scoring chances and as a result gave up more chances.

    Earlier in the year, I compared scoring chances of the Leafs after the first 40 games under Babcock to the final 40 games of the Leafs under Carlyle; Carlyle’s tenure with the Leafs ended after 40 games of the 2014-2015 season. I believe these two periods are worth considering in an assessment of Carlyle because the two teams during the noted time frames were largely the same, except Kessel was not on the Babcock team.

    Compared to Carlyle’s club, Babcock’s team generated more scoring chances and gave up fewer scoring chances at even strength; Babcock’s team were much better at suppressing scoring chances on the penalty kill. Below is the data from

    At 5v5 play, the rankings for Scoring Chances For Per 60, Scoring Chances Against Per 60, High Danger Scoring Chances For Per 60, and High Danger Scoring Chances Against Per 60 are as follows:

    Carlyle: 11; 30; 5; 30 –
    Babcock: 3; 16; 3; 16.

    In terms of the power play, the two teams were similar at generating scoring chances.

    Looking at the penalty kill, the rankings for Scoring Chances Against Per 60 and High Danger Scoring Chances Against Per 60 are as follows:

    Carlyle: 26; 22 –
    Babcock: 5; 1.

    One of the contributions of a coach is the impact that he has on the way the team plays. The Babcock-coached team, compared to the Carlyle-led Leafs, produced better results in terms of creating scoring chances for and also suppressing scoring chances against. This conclusion is supported by both the data and the eyeball test. Therefore, Randy could have done better with what he had and his bargain, if he indeed consciously made that bargain, to generate more scoring chances for in exchange for giving up additional scoring chances against was not necessary.

  • silentbob

    Two things can be true at the sametime.

    Carlyle had a much worse team with worse players who didn’t play as well as the players that Babcock had. He is ALSO a much worse coach then Babcock.