NHL: Using Expected Goals to find regression candidates for 2017-18
One more sleep until hockey.
Rejoice! Your old friend, the NHL, managed to make it through another offseason without doing something stupid like having another lockout. But before the puck is dropped in Winnipeg to kick off another six-month journey, let’s take a look back to help us going forward.
All the stats for this article are provided by Corsica. There are a lot of good websites for advanced hockey stats, but the work done by Manny Elk at Corsica is incredibly important to all hockey nerds, fans and handicappers. If you are betting your hard-earned money on hockey, make sure that you’ve got Corsica bookmarked.
Expected goals (xG) (or expected points in other sports like NFL) is exactly what it sounds like. Not all shots are created equal, and some have more value than others. For example, a wrister from the blue line has much less a chance of tickling the twine than, say, a shot off a rebound from just a few feet away from the net. There is more to it than that, but for the purposes of this article we can keep it simple.
Knowing that means you can figure out what Expected Goals Plus-Minus (xG +/-) is. You take the Expected Goals For (xGF) and subtract away the Expected Goals Against (xGA). So if Detroit has four xGF and six xGA in a given game, they will have an xG +/- of -2.
Like with all advanced stats, xG tells part of a story but not the whole thing. There are a couple of important things to remember when evaluating teams using xG — shooter and goaltender talent and luck are not part of the equation.
It should also be noted that for the purposes of this article, we are only looking at 5-on-5 stats. Power play and penalty kill numbers do have their value, but the majority of the game is played at 5-on-5 and that’s where we’re living today.
With the under-the-hood stuff out of the way, we can look at last season and nominate some positive and negative regression candidates.
Positive Regression Candidates:
Last year, the Boston Bruins led the entire NHL in xG +/- with 29.7. According to this model, the Bruins should have scored nearly 30 more goals than they gave up. In reality, the B’s finished with a 5-on-5 goal differential of -4. It was an odd season from the Bruins. Their underlying numbers were elite, but they were incredibly unlucky. They are an obvious candidate for positive regression after their 95-point season in 2016-2017.
Coming off a historically bad season in 2016-17, the Colorado Avalanche can only go up, and according to their xG +/-, they should. Last year, the Avs finished with an xG +/- of -26.87, which was the second-worst mark in the league. Their actual goal differential was -80. So, yeah, the Avalanche were really bad last season, but they were also on the unfortunate side of the ledger.
After years of dominance, the Los Angeles Kings took a step back in 2016-17, leading to a regime change in Hollywood. Even with tempered expectations coming into the new year, the Kings are hard to write off. Their possession numbers last year were the best in the league, and their 24.41 xG +/- was the third-highest in the NHL behind Boston and Minnesota. Despite those numbers, the Kings finished the year outside of the playoff picture sporting a -7 goal differential at even strength.
Rounding out the positive regression candidates are the young and exciting Carolina Hurricanes. The Hurricanes have been a positive possession team in each of the last three seasons, and in 2016-17 the Canes finished with an xG +/- of 6.46. Unfortunately, Carolina’s goaltending was horrific, and they weren’t able to turn their opportunistic style of play into results, leading to a -24 goal differential and leaving them out of the playoff picture. This offseason, the Hurricanes went out and acquired Scott Darling to stabilize their goaltending, and they signed Justin Williams to help turn those expected goals into actual ones.
Negative Regression Candidates
Two-time defending Presidents’ Trophy winners, the Washington Capitals, put together another stellar regular season last year, but once again fell short in the playoffs. Washington has a very good goaltender in Braden Holtby and a host of talented shooters, so it isn’t surprising to see their xG numbers the way they are. Last year, the Caps finished with an xG +/- of 8.27 and an actual goal differential of 66. Washington will be good again in 2017-18, but a third-straight 55-win season is probably a stretch.
The Columbus Blue Jackets were the surprise team of the NHL last year, largely due to the herculean efforts of their goaltender, Sergei Bobrovsky. Last year, Columbus finished with an xG -0.03, but their actual goal differential at even strength was 36. Like Washington, most pundits expect Columbus to be in the mix for a playoff spot again in 2017-18, but they were probably flattered a bit by their 108-point effort last year.
Along with the Kings, the Chicago Blackhawks have been the torch-bearers for the Western Conference for a while now. Last year, the Hawks put together a somewhat surprising 109-point campaign and won the Central Division title. Chicago has a wonderful goaltender and a lot of talented offensive players, so it isn’t surprising to see them on this list, but their -5.71 xG +/- compared to their actual goal difference of 28 is food for thought.
As is the case every year, the New York Rangers seem a prime candidate to fall out of the playoff picture. Last year, the Blueshirts finished with a -7.2 xG +/- but were plus-14 in the real world despite a down year for Henrik Lundqvist. The Rangers were also a poor possession team, but still found a way to collect 102 points. It’s becoming a tradition to claim that this is the year that the Rangers will fall off, but it’s hard to be confident that they will improve given their underlying numbers.