“Expected” metrics are a popular way of understanding how a team’s under-the-hood numbers might probabilistically manifest as more conventional metrics. That is to say it uses the NHL’s play by play, which captures information about shot type, distance, and angle, among other variables, to add a little more depth to pure volume based metrics, like Corsi. You can read more about it here, if you’re into that sort of thing.
The defending Stanley Cup Champion Washington Capitals are currently 6th worst in the League as measured by their expected goal share, registering a 46.4 xGF%. To wit, based on the volume and quality of shots they’re taking at 5v5 as compared to the volume and quality of shots they’re allowing at 5v5, the model thinks the most probable result is that the Caps should have taken 46.4% of the 5v5 goals scored in their games to date.
This is generally held up as an indication that the Capitals aren’t a particularly good team, and are skating by primarily on hockey analytics’ controvertible notion of “luck”.
Reality vs expectation— Sean Tierney (@ChartingHockey) November 24, 2018
Good teams that are good: TBL, MIN, PHI (cc: @MathHappens51, @Kate_P31), BOS, CBJ, MTL
Good teams that aren't actually that good: NYI, COL, WSH, TOR (uh-oh)
Bad teams that should be better: VGK, LAK, EDM
Bad: DET, VAN, OTT, CHI, FLA pic.twitter.com/CNbmP6qMlb
The root of the problem here is that these expected models generally don’t account for individual shooter talent — something the Caps obviously have in abundance, and have had so for a long time. Here’s a quick look at each team’s xGF% versus their actual GF% dating back to 2007, which is as far back as this data, from Corsica, is available. A tip on reading this, if the solid line is above the dashed line, the team was outperformed their expected goal share for that year, and vice versa.
You can see that this passes a high level sniff test. The expected time series generally resembles the actuals. However, there are patterns that probably can’t be explained by “luck”, such as Carolina consistently underperforming their expected metric.
Here’s a look at the difference in percentage point that each team has averaged over this period.
So, over the past 11 years, only Anaheim has averaged a greater margin above their expected goal share than the Washington Capitals. This shouldn’t come as a surprise to fans who keep an eye on the underlying numbers (or who have frequented this website for many years). The Capitals have rarely been a strong possession team, even through seasons where they’ve won a lot of games. However, the Caps have been fortunate to maintain a strong offensive core for the majority of this time period, which is not common for teams around the League.
So let’s take this down to the last six seasons, including this year, which will more closely resemble the lifespan for a certain quality of play for a team, given roster attrition over time.
The tendency for the Capitals to outperform their expected goal share is even more pronounced over the last five complete seasons, and so far in this one. More so, in fact, than any other team, and by an entire .5 percentage points at that.
It’s easy to point to guys like Alex Ovechkin, Evgeny Kuznetsov, Nicklas Backstrom, T.J. Oshie, John Carlson, as well as former Caps such as Mike Green and Marcus Johansson, as accounting for this gap for being high-skill guys with above average shooting talent. But as with all share metrics, there are two components, so let’s break it down.
Here’s a look at the team’s scoring compared to expected scoring.
The Caps’ scoring talent is obviously contributing to their “overperformance”. Now, how about the goaltending?
So, all told, in 3 of the past 5 completed seasons, and again this season, the Capitals are allowing fewer goals than the model thinks they should, while in all 5 of those seasons, and again this season, scoring more goals than the model thinks they should. That’s not luck. That’s just a team that’s better than a model thinks it is.