I say this a lot but I sometimes feel like nobody hears me: a player’s Corsi% can’t be divorced from the context in which he plays.
There’s been some talk lately about re-signing Luke Gazdic. I’m opposed to this but I feel compelled to note that I think people overstate the case against him (or, indeed, any fourth line scrapper). From my perspective, although he appears to be elite at punching large mammals in the face until they fall down, he isn’t an NHL calibre hockey player. I think he makes the Oilers fourth line substantially worse when he’s on the ice than they’d be if they decided that they didn’t need to carry a heavyweight. The data is there at stats.hockeyanalysis.com for anyone to see – I’d guess that there’s a five point Corsi penalty that comes with playing Gazdic on the fourth line over a generic NHL fourth liner. If the Oilers were a better team, the penalty might be even bigger.
We care about Corsi% because it predicts future goals for/against better than just using goals for/against. We care about goals for/against because that’s what wins hockey games. To me, it’s pretty much inarguable that Gazdic, just like other players selected for their punching generates significantly worse results than the fourth line players who aren’t in the NHL because they can fight. We’ve seen a zillion teams talk about how their fighter can be a hockey player and it basically never comes off. Gazdic is 24 years old and, while it’s admirable that he’s working to be a better hockey player, he almost certainly is what he pretty much will be. If you’re playing him, you’ve made the calculation that whatever gain you get from his fighting outweighs the price of putting him on the ice.
There’s a caveat to this though. While Gazdic and his ilk tend to get slaughtered during the hockey playing part of a hockey game, it’s probably less important than the Corsi% or the goal difference would suggest. I did a post about a year ago talking about the importance of context when talking about Corsi% and made this point:
I suspect that what I call the “win value” of the goals that the fourth line tends to be involved with are lower too. Fourth lines tend not to play in the last ten minutes when the game’s on the line. I would think that a higher percentage of the goals in which they are on the ice are irrelevant to the outcome of a hockey game than with guys who play higher up.
The “win value” concept isn’t that complicated. Some goals are worth more than other goals. Consider two evenly matched teams, each with a 50% chance of winning a game. Now consider two scenarios. First, Team A gives up a goal five seconds into a game with the score tied 0-0. Second, Team A gives up a goal with five seconds remaining in a game with the score tied 0-0. The goal in the second scenario has a much, much bigger “win value” in that it will take the chances of Team B winning from 50% to close to 100%. In the first scenario, there’s a bump, sure, but the odds might go from 50% to 70% – I haven’t actually calculated this but you get my point. A Luke Gazdic error on a goal at the five minute mark is a lot less fatal than a Luke Gazdic error on a goal at the 59 minute mark.
I’ve vaguely intended to illustrate how coaches use their fourth lines, particularly fourth lines with guys who are primarily fighters, for a while. With the chatter about Gazdic lately, now’s as good a time as any. I went through the Oilers games and generated when Gazdic was on the ice at 5v5 at certain points (based on time and score) and when Jordan Eberle was on the ice at 5v5 at certain points. Eberle’s kind of at the opposite end of the scale from Gazdic and has played every game this year.
So here’s score tied. I’ve basically broken each period into four five minute parts. Overall, Eberle gets about 2.9 times as much TOI with the score tied as Gazdic does. You can see that the bulk of Gazdic’s ice time is occuring earlier in tie games. When the game is tied in the first 35 minutes, Eberle gets about 2.5 times the TOI that Gazdic does. When it’s tied after that, Eberle gets about 5.5 times the TOI that Gazdic does. When the game is tied in the final 15 minutes, Eberle gets about 9.8 times the 5v5 TOI that Gazdic does.
Let’s look at one down and one up. Eakins has talked a lot this year about wanting to find more minutes for his fourth line generally. He’s frequently mentioned that it’s harder when the Oilers are trailing a game because you want to give your skill guys a chance to get you back in the game but that it’s easier when the Oilers are leading.
Overall, the data backs up that he does what he talks about – Eberle plays 4.3 times as much as Gazdic when the Oilers are down a goal and just 2.2 times as much when they’re up a goal. Focusing on the down one side of the ledger, it’s interesting to me that Eberle’s ice time relative to Gazdic spikes in the second half of the period – with a TV time out at 10 minutes and again at six minutes as well as a looming intermission, I’d guess you can ride the horses a little harder. Down one in the third, the whole period gets nuts. Eberle plays 12.3 times as much as Gazdic.
The up one side is kind of neat too. I found it sort of interesting that Gazdic’s played more 5v5 TOI than Eberle has up one between 5:01 and 10:00 of the first. I would bet if I looked, I’d find that the Oilers were most likely to be leading off an early PP goal, which puts Gazdic in a position to get some TOI as the PP unit rests. There’a slight bump in Eberle relative to Gazdic as the second winds down and then a real spike in the last ten minutes of the third, where Eberle has played a smooth 153.6 times as much as Gazdic with the Oilers protecting a lead.
The first ten minutes of the third would catch my eye if I was an opposing coach. You’d have to look at it with a little more detail but Eakins really does seem comfortable playing his fourth line in that circumstance – the TOI gap between the fourth line and Eberle has been pretty small. If I coached another team, I might explore that a little further, look whether it happens on the road in addition to at home, that sort of stuff. It gives you a bit of a window to take your shot at the fourth line when you’re chasing the game that you might not get later in the third. (As ever, sample sizes are small. If I was trying to build a dossier on a coach, I’d look at his whole career.)
What about two down and two up?
There’s a bit of a push in the third when trailing by two, in terms of bumping up Eberle’s TOI relative to Gazdic’s but not nearly as much of one. Truth is, a game where you’re down by two is virtually certain to be lost. Up two, it’s the time before the intermissions and the last ten minutes that catches my eye. That seems to be where Eakins really turns the screws on the other team – get to the intermission/end of the game without conceding.
Finally, garbage time, when one team or the other is up more than three goals.
It’s eye catching that when the Oilers are down three or more, there’s a bit of a push with Eberle right before the second intermission. If the game’s in the third and the Oilers are out of it, Gazdic plays just as much as if it’s the first period of a tie game. When the Oilers are blowing up the opposition, Gazdic plays quite a bit relative to Eberle – he actually has played more 5v5 TOI in the last fifteen minutes of a game this year with the Oilers leading by three or more than Eberle has.
What does this all mean? Well, I think it’s fair to say that Gazdic’s ice time is different from Eberle’s in that it tends to take place earlier in the game, when a goal has less of a determinative impact on the outcome or late in the game in a blowout, when a goal has little impact on the ultimate result. If those things are true, the Corsi% when he’s on the ice is somewhat less important than it might otherwise be because the average goal scored when Luke Gazdic is on the ice is less relevant to the outcome of the game than the average goal scored when Eberle’s out there.
Our usual rule of thumb is that three goal difference equals a point in the standings. While that’s true on the team level, one of the implications of this is that it may not be true on a line level. Assume that you play on a line that scores 3 5v5 goals per 60 and allows 2 5v5 goals per 60. That’s going to translate into more wins if you’re playing Eberle’s TOI, which includes a bunch of crunch minutes, than if you’re playing Gazdic’s minutes, which are slanted towards time that doesn’t matter.
In the Corsi and Context post linked above, I found that the difference between the best and worst first liners was about 20 goal difference. For the best and worst fourth liners, it was about 10 goal difference. It may well be that, given the minutes they play, it takes two goal difference from a first line to equal a standings point and five from a fourth line. I haven’t quantified this but there’s obviously going to be some sort of an effect like this at play.
I wouldn’t go so far as to say that this excuses Gazdic’s presence on the roster or that fourth lines are meaningless. The goals for and against that Gazdic’s on for still happen, he’s going to have a worse goal difference than the typical fourth liner and some of those goals are going to be meaningful. It’s just that the effect is smaller than the goals for/against would suggest though.
A second, potentially more critical issue, is the impact of a player like Gazdic on the other players on the team. We know that hockey players get fatigued, just like other athletes. There’s an interesting passage in a book I read last summer about soccer called The Numbers Game which discusses this:
To compound matters, players do not usually want to be removed unless they are injured. They are experts at making a manager believe they have plenty more to give, which makes judging their performance levels even more tricky…
…another study coauthored by Carling that looked at players’ work-rates after a teammate had been sent off shows that soccer players are experts at pacing themselves and operating at less than full capacity.
As a result, fatigue may not show up in their average workrate, it may not be visible from the dugout. Instead, it shows up when they try to go from the 90 percent capacity they are operating at to the 95 percent they need to stretch for a tackle or to leap for a header. Early in the game, they can do this easily. Later, when they are pacing themselves, reaching the required capacity is no longer possible.
Hockey’s obviously not perfectly analogous to soccer, even if it’s more alike than a lot of people realize. That said, hockey’s substitution patterns seem a little haphazard, with teams all over the map in terms of how much they play players. I suspect that there is an impact to playing a guy twenty minutes a night versus playing him sixteen minutes; it’s just a question of whether any degradation in play is outweighed by playing your Jordan Eberle instead of your Luke Gazdic.
These aren’t necessarily easy questions. A lot of them kind of require blending hard data with your own subjective judgments about things. If it was me, I’d have a fourth line full of guys who were hockey players first and I’d invest some resources into investigating things like whether teams that have fourth lines that can play, like Boston, get better results from their first, second and third lines later in games. I suspect that they do. I don’t think that I’d give away something in terms of on-ice performance to have a guy who can knock people out.
Evidence is evidence though and if you’re going to have this discussion, you have to acknowledge what the data says. The impact of having Gazdic on the roster, in terms of wins and losses related to the goals for and against when he’s on the ice, is smaller than those of us who pay attention to the data would think.Email Tyler Dellow at email@example.com