Deviations: Part I
Some statistical nuggets to tide you over until Saturday’s game.
This is the beginning of a new segment here on Knights On Ice entitled Deviations, which will be a semi-regular aspect of our site focusing on answering questions about the Vegas Golden Knights through the lens of analytics.
Well folks, we have not had a Vegas game in a while, but you’re probably all riding high, especially if you’ve recently hit up Essence on the Strip.
With no game for over three whole days, you’ve needed your fill of Vegas Golden Knights hockey, and I’m here to provide you with something, well, quite short of that. As the resident stat guy here — alongside Mike Hannah — I’ve been digging up some things that may be of interest to you fans out there.
Here are some questions I aimed to answer:
What should the Golden Knights’ record be?
The Vegas Golden Knights have been equal parts surprising and successful this season. That is to say, very. However, the team has not been fueled by cheap wins and off-the-charts goalie play. The Golden Knights have a save percentage of .907, a bit below the league average of .912, while their shooting percentage is admittedly quite high.
But let’s run a quick thought experiment, for which we will primarily use expected goals. Expected goals are just what they sounds like—how many goals a team or player is expected to have based on their shots. The data gets very granular—it factors in the hand of the shooter, the location on the ice, the type of shot (snap, backhand, slap, deflection, wrist) and whether it is an even-strength, power-play or short-handed situation.
Taking a look back at every game, and strictly going by expected goals for each team after 60 minutes, we see the team “should” have a record of 20-13-0, rather than 22-9-2. That configuration results in a drop of six points, and would put them in a tie with the San Jose Sharks for second in the Pacific division, and still the owner of the playoff spot, albeit a more tenuous one.
Some notable games that switched: Three of the four victories vs. Arizona would be erased, while three rough Maxime Lagace performances (10/30 vs. NYI, 10/31 vs. NYR, 11/6 vs. TOR) would change into wins.
For another estimate of how many wins the team should have, we can look at Hockey-Reference’s Point Shares, a metric applied on both the individual and team level that serves as an estimate of how many team points each individual contributed to the team. The Golden Knights have a team total of 41.9, which we will round to 42, led by Jonathan Marchessault’s 3.8.
These two looks at their performance would suggest that they have overperformed by a considerable margin (10-15 points across an 82-game season).
However, we should also keep in mind that the Knights have played nearly 80 percent of the season without their No. 1 goaltender, Marc-Andre Fleury, so the regression may be balanced out by improved goaltending. After all, although Lagace played an important role for this team in the face of compounding injuries, he did allow 15.23 more goals than the average goaltender during his span with the Knights.
How good is the n’th line compared to other lines in the NHL?
Let’s square away things a bit. A hockey team rolls four lines; there are 31 teams in the NHL. So, for the sake of our analysis, we will look at the top 124 lines as ranked by 5-on-5 ice time. As a result, the 16th ranked line overall would be the “average” first line, 47th ranked for second, 78th for third and 109th for fourth.
Everyone still with me? Good.
Again, we are going to take a turn into expected goals as one of our ways to analyze the team, but with somewhat of a controversial twist—we will be looking at xG plus-minus at 5-on-5. This avoids some of the main pitfalls of using +/- (short-handed goals against, too much reliance on teammates, fluke goals, etc.), while retaining what it is we really want to know: how much each of the Vegas lines produce offense and prevent opposing offense.
From the above chart, you see that all of Vegas’ lines rank somewhere around average, with the first, third and fourth entries slightly above and the third a bit below. Here’s where things get interesting, however. Vegas’ worst performing line is one that many would put on par with their top-performing line. The David Perron-Erik Haula-James Neal line is often considered a 1B to the 1A that is the Reilly Smith-William Karlsson-Jonathan Marchessault line, however their expected goals data varies wildly.
Interestingly enough, for a “fourth” line, Vegas’ group does quite well, as does their third line. Unfortunately for the latter trio, they have not turned theoretical offense into real offense in a meaningful way, and their PDO (on-ice shooting percentage + on-ice save percentage) is the 12th lowest in the league. And lest you think that PDO is a function of skill rather than luck, consider that even lower than the Brendan Leipsic-Cody Eakin-Alex Tuch group lies a line featuring Sidney Crosby and another that has Joe Thornton and Joe Pavelski. Sometimes you’re just snake-bitten, and no one fits that bill more accurately than Leipsic.
There are other methods I considered, ranging from simply Corsi For Percentage to Score-Adjusted Fenwick, but I believe this to be the best option that remains publicly available. It can be found for no charge at Emmanuel Perry’s Corsica.
Finally, once you unpack the expected goal differential into goals for and against, it paints a rather fascinating picture:
Expected Goal Data for Vegas’ Four Lines
|Line||Expected GF/60 Rank (Actual)||Expected GA/60 Rank (Actual)|
|Smith-Karlsson-Marchessault||15th (3.17)||40th (2.07)|
|Perron-Haula-Neal||117th (1.64)||44th (2.09)|
|Leipsic-Eakin-Tuch||34th (2.91)||79th (2.35)|
|Nosek-Bellemare-Carrier||72nd (2.40)||77th (2.35)|
Who would have thought that a line with two offensive minded wingers like Perron and Neal would in fact be awful offensively (at 5-on-5 anyway) and quite good on defense?
Bill James, the sabermetrics pioneer once posited that a good metric supports common sense 80 percent of the time and surprises you the rest of the time. Judging by the above table, I’d have to say we’re probably looking at a very good metric. And it is true that expected goals are far superior in their predictive and descriptive power than mere shot data. Still though, it is not perfect, and should never be taken as law.
Is this a playoff team?
Short answer, yes.
Long answer: It’s very likely. According to several respected projection models, the Golden Knights are 76 to 95 percent likely to make the playoffs. It’s why “yes” isn’t the best answer, albeit the most palatable one to Golden Knights fans.
Think about it like this—the Golden Knights have 46 points in just 33 games. Even if they play point-per-game hockey, which is to say quite mediocre, the rest of the way, they are almost guaranteed a playoff spot, as a 95-point team is a virtual lock to make the postseason.
Since the division realignment that began with the 2013-14 season, 31 of 34 Western Conference teams with at least 90 points made the playoffs, and just one of those (the 2014-15 Los Angeles Kings) finished with as many as 95 points.
While it is unclear what the trade deadline holds — namely whether or not general manager George McPhee sticks to the three- or six-year plan — it appears that Vegas is far more likely than not to play beyond April 7th.
Stay tuned for our next edition of Deviations, which will look at things like Brendan Leipsic‘s effectiveness, optimized defensive pairings and what it means to be a part of the Vegas “fourth” line.