Last week we reviewed the advanced statistic “expected goals for”. It’s a metric that analysts will utilize quite a bit this season, so we thought it might be beneficial to provide a bit of an introduction to the stat for those not familiar. If you are still a little fuzzy about the stat, or need a refresher, here’s a quick recap.
Expected goals for is a statistic based on the aggregation of each exact shot attempt location and type of shot, and the historical success rates for those shots/types taken at each location.
For example, if Alex Ovechkin takes a shot from the top of the right circle, and historical data for that exact location says 5% of the shots have been goals, Ovechkin would be credited with a 0.05 expected goals for (xGF) value for that one shot attempt. The xGF values are summed over the course of a game or other time segment, yielding a resultant “expected goals for” (xGF) value. If Ovechkin made five shot attempts in a game with an aggregate expected goals for of 2.35, he earned an xGF of 2.35.
The same is calculated for the opposition while Ovechkin is on the ice. If an opposing player takes a shot that has a 3% historical value of scoring, while Ovechkin is on the ice, the opposing player is credited with the historical value of 0.03 xGF and Ovechkin is credited with an expected goal against (xGA) of 0.03.
We can sum the player’s expected goals for and expected goals against to achieve an overall expected goals differential, which would be 0.05 – 0.03 = +0.02. Anything in the positive conveys that the player is generating more scoring opportunities than the opposition when the player is on the ice.
2022-23 Preseason Values
Here is a look at the Capitals players and their expected goals differential (xGF – xGA) for the 2022-23 preseason, so far. [Click to enlarge].
There are no real surprises at the top of the list (the best differential). Lucas Johansen has the best differential among all prospects in camp that have played in a game. In other words, the Capitals are generating more shot attempts/better quality of scoring chances than the opposition when he is on the ice. Joe Snively and Hendrix Lapierre are a close second and third among prospects.
At the low end, Dylan McIlrath is really no surprise, as his brand of defense is a stay-at-home style. Connor McMichael has the lowest differential among all prospect forwards.
CAVEATS AND CONTEXT
As with ALL statistics, the expected goals differential metric is a single brush stroke to an overall painting. It provides part of the overall picture, but in no way should it be used alone to render an overall verdict or opinion on a player.
The current expected goals differentials presented above includes a number of significant variables. Players have played a different number of games, with varying linemates of varying expertise, against a wide range of opposing expertise.
By Jon Sorensen
That’s a weird stat.
Ovi of 10 years ago shooting from the same office spot is likely to be a better chance rate of success than Ovi in the same spot in 2022. That makes sense to me, does it to you?
How about this one. McM’s chance of scoring in a location deemed 10% xGF might have the same success rate as an Ovi shot from a 5% xGF location.
What this tells me, is that although this stat may be interesting to coaches (who could probably tell this through observation rather than stats) who might use the info to coach his team, it doesn’t provide a reality of what real xGF expectation’s really are. The player makes the difference. Who is shooting that puck. Whether for or against.
The expected goals model accounts for what you mention.
If you are interested:
https://www.nhl.com/kraken/news/analytics-with-alison-expected-goals/c-327728890
Thanks, I’ll do that. I fight my resistance to learn various stat and eval changes that have taken place in NHL and MLB. It’s hard for an old dog to learn new tricks sometimes.
You are not alone. In fact, you are in the vast majority when it comes to hesitance regarding new stats.
We have made a concerted effort to present these new “advanced stats” in basic form, at the start of each season, in hopes of bringing a few more along for “the ride” each season. Some will, but most won’t, and that’s ok, too.
Do I have this correct? It is not just that player’s goals but team goals? And it doesn’t take into account which other Caps are on the ice at with him? So someone who is bounced around from position to position, line to line as a sub would probably not do as well?
It is quantified for all skaters when a player is on the ice.
I’m not sure I follow second part of your question.
I would expect an improvement from a player if on ice with a team’s stars rather than being on the ice with players disposed of at the trading deadline. Teammates count.
Same goes for a player who knows his linemates well and has played with them (that is the reason I think the rookies did so well last October — we had the Hershey Caps on the ice).
Likewise, a player who is a sub filling in different positions and never getting comfortable in one spot is probably not going to be as productive. When I was a sub teacher I think I did well, but I did better when I repeated in the same job.
Just some things that I think could be asterisks to a player’s score.
Of course, there are always injuries which we know nothing about too.
Your general overall assumption is correct, however, it doesn’t always result in the stars having great xGF%. Here are the xGF, xGA, xGF – xGA and xGF% for each of the Capitals skaters last season at even strength (5-on-5). Notice Ovi less than 50% (more xGA than xGF)