Ranking the top 10 lines in the NHL

There’s a lot of noise surrounding lists these days, we are still in August after all, and determining the best lines in the NHL is a topic that hasn’t had too much chatter around it. So, that’s what we’ll be doing today.

Just what makes a line the best, most effective, and most impactful line in the league can be a lot of things. For me, it’s a line’s ability score and drive play at a high rate, out-score their opposition, and their ability to give their team powerplay opportunities while staying out of the box themselves.

Eligibility

Rather than just throwing names out there, we should really begin by narrowing the list. These are supposed to be the best of the best after all. First things first, we’re talking about actual lines, not just three players that happen to be on the same team. For example, no combining Sidney Crosby and Evgeni Malkin’s points, nor will we be doing that with John Tavares and Mathew Barzal.

Secondly, something that ties together with the first point of eligibility, the statistics used from this point on will be from 5-on-5 play only. Using power play production to determine the best lines in the NHL just doesn’t make sense. Also, this feels like a good place to mention that any and all shot metrics used will be score adjusted unless specified otherwise.

Finally, there has to be some sort of minutes cutoff. Sure, it’s undeniably great for a line to have a 61.61% Corsi for and 70.33% expected goals for, but if they only played just over 147 minutes together, the sample size just isn’t there. Looking at you Rick Nash, Mika Zibanejad, and Mats Zuccarello. After mulling over the options, it was decided to go with a minimum of 300 minutes at 5-on-5. Now, this is a fairly arbitrary number, but what it does do is help cut down the number of eligible lines to just 44.

With that out of the way, let’s begin finding the candidates.

Production

Let’s ease our way into this by starting with the most straightforward stat out there, goals. Scoring more goals than your opponent is good - shocking, I know! On the left we have the trios that scored the most goals this season and on the right we have those who had the best on-ice percentage.

Now the forwards aren’t entirely responsible for their on-ice goal results. For example, the Adam Henrique line benefited from a 97.27 on-ice save percentage, much higher than the league average of just above 92%. Goaltending, combined with the quality of defenders that a line plays with, can have a huge impact on a line’s goal numbers. So can luck, but to put it simply; this is what happened when these lines were out there, and that counts for something.

Then there’s the rate stats. Ten different lines stand out for either scoring at a high rate or being scored on at a low rate last season:

Again, a good amount of this falls on goaltending, as was can see by the Henrique line once again taking one of the top spots, but this is also where some of the more effective lines stand out offensively. From the Flyers perspective, it’s some pretty good company to see the Sean Couturier, Claude Giroux, and Travis Konecny line with.

Shot Metrics

The next thing that we’ll be looking at are Corsi and expected goals. If a line is consistently out-shooting and out-chancing their opposition, they’re clearly doing something right on the ice. Leading the way here is the Winnipeg Jets’ line of Adam Lowry, Andrew Copp, and Brandon Tanev. While the trio certainly lacks the star power of the previous lines discussed, they were the only line of the 44 eligible to surpass the 60% mark in both Corsi for and expected goals for.

Minnesota’s line of Mikael Granlund, Mikko Koivu, and Jason Zucker stand out here for having the largest discrepancy between the two stats. This isn’t just something that one line did for the Wild, they as a team finished the regular season with the sixth worst Corsi for percent while also boasting the best expected goals for percent in the league. Only one other team that finished in the top half of the league in xGF% had a CF% below 50 and - you guessed it - it was the Flyers.

Shifting back to the trio, If we look at the un-adjusted numbers for a second, we’ll see that they had a CF% of exactly 50; 367 for, 367 against. That’s strange enough by itself, but even then their xGF was still well above 64%. That’s Wild.

Both statistics should still be viewed as separate entities, and the average of the two shouldn’t be seen as anything more than a easy way to see which lines excelled in both unweighted shots and weighted shots throughout the season.

Penalties

Finally, the last piece of the puzzle, penalties drawn and penalties taken. There’s really not much to say here, putting your team on the powerplay is a good thing and directly leads to better scoring opportunities and more goals scored. These trios were the most proficient at both drawing penalties and not taking many themselves.

It’s no surprise to see Lady Byng finalist Aleksander Barkov and his linemates at the top of this list, and with that award in mind, you’ll notice the lack of William Karlsson. He may have not taken many himself, five to be exact, but he only drew nine penalties at 5-on-5. That, combined with how his linemates Jonathan Marchessault (-3) and Reilly Smith (-7) weren’t exactly the best in this area of the game, led to the line’s -9 penalty differential.

Final Ranking(s)

There was a lot of data discussed and going about how to weigh each data point was something that I knew would be hard going into this. The quickest and easiest way to come up with an answer would be to, well, not weigh them differently at all. If you take every line’s rank in every category above and average the results, the top-10 would look like this:

1. Fiala, Smith, Turris (NSH)
2. Couturier, Giroux, Konecny (PHI)
3. Barkov, Bjugstad, Dadonov (FLA)
4. Schenn, Schwartz, Tarasenko (STL)
5. Couturier, Giroux, Voracek (PHI)
6. Copp, Lowry, Tanev (WPG)
7. Atkinson, Dubois, Panarin (CBJ)
8. Karlsson, Marchessault, Smith (VGK)
9. Arvidsson, Forsberg, Johansen (NSH)
10. Hyman, Matthews, Nylander (TOR)

However, I decided that was a bit too simple for my liking. It didn’t feel right to have every stat used be worth the same as one another, so with the goal of getting offensive production to count for more, I turned to a stat that I’ve been a fan of for awhile now; Game Score.

Back in the Summer of 2016 Dom Luszczyszyn introduced Game Score to the hockey world with his “Introduction to Game Score” piece over on Hockey Graphs. His goal was to find a way to measure single game productivity and present it as a single number for every player that participated in said game. In my opinion, the thing that makes Game Score great is it’s simplicity, and that’s the exact vibe I wanted to recreate here.

Formulas aren’t inherently easy to understand by any means, but, like Game Score, I feel like what I’ve come up with is pretty painless to grasp. As mentioned earlier, I wanted to make goals worth more - that was the goal in mind. With lines in the NHL being as fluid as they are, I decided that I wanted this stat to be as descriptive of last season as possible. In essence, I was going to value goal metrics over shot metrics, also known as why Robert Hagg played in 70 games this past season while Travis Sanheim wound up in the AHL. No, I will never stop talking about that. Anywho... here’s the formula I settled on for this exercise:

(Goals * 0.6) + (Goal Differential * 0.3) + (Corsi Differential * 0.1) + (Expected Goals Differential * 0.1) + (Penalty Differential * 0.1)

If you’re well versed in Game Score knowledge you’ve probably already noticed the stats that I’ve either A) omitted or B) added. Oh, and that the values were changed. More on that later. The first thing I did was remove assists from the equation, reason being that assists just simply aren’t compatible with trios. Each line is already being credited for every goal scored while they’re on the ice, so assists are irrelevant. Blocked shots and faceoffs were also removed from the equation due to how little they impacted Game Score and, again, because they just don’t work well with trios. Now, the final thing that I removed was shots on goal. This was a last minute judgement call that I made based on two thoughts; I wanted to keep goal differential worth half of what goals were, and also wanted to keep goal differential worth more than the two shot metrics combined. Removing shots on goal was the only way to keep both of those statements true without changing all of the numbers, which could have been done, but having three factors surrounding shots seemed a bit unnecessary anyway.

In Game Score goals are worth .75 but with more players involved comes more goals involved. This made goals a bit too overbearing and, rather than scale the rest of the numbers up, it was decided to scale goals back by .15 to keep the final outcome a smaller number. Since I was introducing expected goals to the equation, I also dropped the value of Corsi by a minor amount and gave expected goals the same value. Finally, I took penalty differential and dropped it’s value as well, by .5 to be exact. This did two things. First, it kept penalties worth the same as Corsi, just like they are in Game Score and, more importantly, it kept goals worth the same amount as everything else combined. All of this turned into what I like to call Line Score.

So with that out of the way, let’s move on to the results. The following lines graded out as the best among the 44 eligible line combinations in both Line Score overall, and Line Score per 60 minutes.

So there you have it, the title of best line in the NHL during the 2017-18 season goes to the Vegas Golden Knights’ line of William Karlsson, Jonathan Marchessault, and Reilly Smith. Way to go, Dale Tallon!

Barkov, Nick Bjugstad, and Evgenii Dadonov may have been the most effective trio in their time together, but the Golden Knights’ top line provided more value over the course of the season than they did.

Closing Thoughts

Is this a perfect method to determine the best line in the league? Nope, absolutely not. However there’s no single-stat out there to judge lines (to my knowledge) and I wanted exactly that; a single number to rank them by. There’s surely more sophisticated ways to do this but this gave me something to use to rank them by, and allowed me to try something new.

One of the cooler things about this exercise is how it brought some of the more under the radar line combinations, at least to me, to the surface like Winnipeg’s Andrew Copp line and Nashville’s with Kyle Turris.

Shifting our focus to the Flyers for a second, since that’s what we do best around here, it’s no surprise that both of the Giroux lines make an appearance in the final results. While I still think the better move going forward would be to keep Konecny on the top line, allowing Nolan Patrick and Jakub Voracek to build on their strong play together during the second half of the season, it’s interesting to note that the top line with Voracek ranks pretty high up there when accounting for minutes played.

Lastly, I want to make it clear that was loosely based on the concept of Game Score. A lot more research went into Dom’s equation and there’s actual repeatability to it. This was just a fun way to rank the top ten lines in the league.

All data used courtesy of Corsica.hockey