Revisiting the expected goals model
By Ryan Pike
8 years agoThe past is a pretty good predictor of the future, which is why a lot of models of future performance are based upon past performance. Earlier this season, we used past shooting (and saving) performances to predict how many goals the Calgary Flames should be expected to score (or allow).
As you’d expect, a lot of the team’s issues can be heaped upon a handful of players when performance is analyzed through this lens.
Let’s dive in!
THE METHOD
We took a look at this back in January. To quote, us, back then:
Admittedly, the Expected Goals model I am using here is simplistic. I
like it because it’s simplistic, but that does limit its effectiveness
sometimes. Here’s the gist: based on a player’s career shooting (or
save) percentage, you can project how many goals they should score (or
allow) based on the number of shots they have (or have faced). This
approach pinpoints individuals that are scoring more or less than they
should be, and goalies that are allowing many more (or fewer) goals than
they should be.All shots are equal under this formula, which is
one drawback. And for younger players – Sam Bennett, for example – it’s
less effective because for them the current season could be their entire
career. It’s a tool, but it’s not bulletproof by any means.
SKATERS
Through 73 games; for difference, a positive number is good. Significant numbers are bolded.
| Player | Actual Goals | Expected Goals | Difference | 15-16 S% | Career S% |
| Gaudreau | 27 | 27.7 | -0.7 | 14.1 | 14.5 |
| Monahan | 26 | 27.9 |
-0.9 | 14.4 | 15.5 |
| Giordano | 19 | 14.5 | +4.5 | 9.9 | 7.6 |
| Colborne | 16 | 12.4 | +3.6 | 18.4 | 14.2 |
| Backlund | 15 | 11.6 | +3.4 | 10.9 | 8.4 |
| Frolik | 15 | 11.7 | +3.3 | 10.4 | 7.9 |
| Bennett | 15 | 14.9 | +0.1 | 12.8 | 12.7 |
| Hudler | 10 | 12.0 | -2.0 | 12.5 | 15.0 |
| Hamilton (D) | 10 | 9.5 | +0.5 | 6.1 | 5.8 |
| Jones | 9 | 9.1 | -0.1 | 13.4 | 13.6 |
| Stajan | 6 | 7.3 | -1.3 | 10.9 | 13.3 |
| Brodie | 4 | 4.0 | 0.0 | 5.8 | 5.8 |
| Russell | 4 | 2.8 | +1.2 | 7.1 | 5.0 |
| Ferland | 4 | 4.6 | -0.6 | 3.6 | 4.1 |
| Jooris | 4 | 7.4 | -3.4 | 5.3 | 9.8 |
| Granlund | 4 | 4.6 | -0.6 | 10.8 | 12.5 |
| Raymond | 4 | 5.1 | -1.1 | 7.5 | 9.6 |
| Wideman | 2 | 4.6 | -2.6 | 2.7 | 6.1 |
| Bouma | 2 | 3.9 | -1.9 | 4.7 | 9.1 |
| Engelland | 1 | 2.9 | -1.9 | 1.6 | 4.7 |
| Bollig | 1 | 1.6 | -0.6 | 2.1 | 3.3 |
| Nakladal | 1 | 1.0 | 0.0 | 3.7 | 3.7 |
| Hamilton (F) | 1 | 0.3 | +0.7 | 25.0 | 7.1 |
| Jokipakka | 0 | 0.3 | -0.3 | 0.0 | 2.5 |
| Hathaway | 0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Kulak | 0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Smid | 0 | 0.4 | -0.4 | 0.0 | 3.3 |
| Grant | 0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Wotherspoon | 0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Agostino | 0 | 0.2 | -0.2 | 0.0 | 6.7 |
| TEAM | 200 | 202.2 | -2.2 | 9.4 | 9.5 |
GOALIES
Through 73 games; for difference, a negative number is good. Significant numbers are bolded.
| Goalie | Actual GA | Expected GA | Difference | 15-16 Sv% | Career Sv% |
| Ramo | 94 | 97.2 | -3.2 | .909 | .906 |
| Hiller | 76 | 55.2 | +20.8 | .883 | .915 |
| Ortio | 40 | 40.5 | -0.5 | .901 | .900 |
| Backstrom | 1 | 1.9 | -0.9 | .955 | .915 |
| TEAM | 211 | 194.7 | +16.3 | .900 | .908 |
THE NUMBERS
Significant Positive Differences (A Goal Or More):
- 4.5 – Mark Giordano
- 3.6 – Joe Colborne
- 3.4 – Mikael Backlund
- 3.3 – Michael Frolik
- 3.2 – Karri Ramo
- 1.2 – Kris Russell
Significant Negative Differences (A Goal Or More):
- 20.8 – Jonas Hiller
- 3.4 – Josh Jooris
- 2.6 – Dennis Wideman
- 2.0 – Jiri Hudler
- 1.9 – Sean Monahan
- 1.9 – Lance Bouma
- 1.9 – Deryk Engelland
- 1.3 – Matt Stajan
- 1.1 – Mason Raymond
Where do we begin? Okay, let’s get this out of the way: it’s probably unfair to single out Jonas Hiller for sinking the Flames season, but if he was anywhere close to what he used to be – or close to what he was last season – he probably gives up at least 10 fewer goals than he did and the Flames are in the playoff hunt. They might not be in, but they’re in the mix.
On the sunny side: Giordano has really gotten hot lately, which has driven his expected goals into the black quite nicely. Colborne’s shooting percentage shouldn’t be the best on the team and is due for a big correction, but given where he usually shoots from (right in front of the net) he should have a percentage around 12 or 13. Backlund and Frolik have benefited from playing with each other, and Ramo was quietly quite good for a while before he got hurt.
On the down side: Already mentioned Hiller (yeesh), and any questions about Calgary’s lack of scoring depth should be directed to the handful of players that haven’t scored at their expected rates. Prime among them are Monahan and the departed Hudler, who were ice-cold for the first couple months of the season.
Recent articles from Ryan Pike
