A data-driven look at performance of Golden Knights’ top line

By Justin Emerson

Thursday, Oct. 17, 2019 | 2 a.m.

If you see a top scorer on a breakaway with no one to beat except a backup goalie who has been shaky all game, you might expect a goal.

It happened last week in Arizona, where William Karlsson got loose on a breakaway and came as close as he could to scoring without putting the puck into the net. When you see Karlsson slide behind the defense with no one to beat but the goalie, you probably expect a goal.

However, that would not be an “expected goal,” according to the hockey analytics community. In fact, Moneypuck.com says that play was worth 0.242 expected goals.

At its core, expected goals measure the space on the ice where a shot is taken and calculates the probability of it going in. It is a stat that has become ubiquitous in the analytics community, including by NHL teams in preparing for an opponent, or analyzing their own performance.

Karlsson and linemate Jonathan Marchessault have combined for one goal this season, though both should have more, using the traditional eye test as well as the expected-goals metric.

There are two ways to measure expected goals. One is individual expected goals (ixG), which is the sum of the expected-goal value of a player’s shot attempts. It’s a counting stat for how many goals a player should have. If he has more goals than expected goals, he could be getting lucky, or vice versa.

Then there is on-ice expected goals, which is a measure of expected goals for and against a team when a player is on the ice at 5-on-5. That is then listed as a percentage where anything above 50% is good (the team has more expected goals with that player on the ice).

As a team, the Golden Knights have generated a league-best 14.18 expected goals at 5-on-5 and have allowed 11.41 expected goals, according to Natural Stat Trick before Thursday night’s game. That’s an expected-goals-for percentage (xGF%) of 55.42, fifth in the league. Their actual-goals-for percentage is 48.48%, indicating that they have actually been quite unlucky this year.

What’s even better for the Golden Knights is the play of the team’s top line, whose three players have some of the best on-ice expected-goals numbers in the league. Among players with at least five games, Reilly Smith is second in the league with an xGF% of 72.65, William Karlsson is third at 71.04%, and Jonathan Marchessault is fifth at 69.69%.

“When you get those chances, it’s positive,” Marchessault said. “We have to keep going in the right direction and it’s about finishing the job when we have some. Those stats are positive.”

What those stats say is that the Golden Knights are driving play with that trio on the ice and should be scoring the majority of the goals. What that stat does not say is who is doing the scoring. For that, we turn back to individual expected goals (ixG).

Marchessault and Karlsson have one combined goal this year in all situations. At 5-on-5, the pair has a combined 3.45 ixG. That Marchessault and Karlsson are underperforming in their ixG is most likely an indicator of a hot goalie, or not getting all of a shot, or what many refer to as bad puck luck.

Marchessault's frustration is evident.

“Well, do they know how many goals Karlsson and I have right now? We have one in seven games combined,” Marchessault said.

Still, he agreed the line is playing well and that the chances are coming. His coach feels the same way.

“I know they get frustrated, the players, because they’re getting chances that aren’t going in the back of the net, but what do you do?” coach Gerard Gallant said. “You keep working, keep grinding and sooner or later they’re going to go in the back of the net.”

Gallant said he gets briefed on what the analytics say about his players and that it helps guide his decisions. The Golden Knights aren’t a team that is going to overreact to a hot or cold streak.

Remember Brandon Pirri’s terrific start to the year when he was called up last year? They saw the same stats we did and paid him close to the league minimum this offseason. He hasn’t scored yet this year and was a healthy scratch the last three games. His xGF% was 40.86 in four games.

And while Marchessault and Karlsson are struggling to score right now, their linemate certainly isn’t. Smith leads the team with five goals, and four of those have come at 5-on-5 with an ixG of 1.19. It’s a terrific start, but regression goes both ways. Don’t count on Smith to finish with the 59 goals he’s currently on pace for.

And don’t expect Marchessault to finish with 12 or for Karlsson to go scoreless. They’re playing quite well, and the goals are coming.

Or at least that they’re expected to.