Today’s column will continue the season-long odyssey into goalkeeper performance metrics, with the recent debate surrounding Craig Gordon’s inclusion as a player of the year finalist and subsequent win making it topical. Of the publicly available data vendors, Statsbomb appears to have the most robust keeper-related data and metrics. So which keeper ‘deserved’ to be a finalist? Joe Hart, Craig Gordon, or a sleeper?
Using traditional keeper-related metrics would place Hart towards the top of the list, as he’s been at the helm for nineteen clean sheets in league games and conceded just 0.47 non-penalty goals per 90 minutes so far this season. By comparison, Gordon has had fourteen clean sheets and conceded at a rate of 0.85 non-penalty goals per 90 minutes. This radar offers a broad array of shot-stopping and goal prevention-related metrics comparing the two:
As we attempt to attribute between controllable and uncontrollable factors, Hart has played for a team which has conceded just 0.52 non-penalty xG (npxG) and faced just 1.85 shots per 90 minutes, while Gordon’s Hearts side has conceded 1.16 npxG on an average of 4.26 shots – both more than double the rates for Hart.
So, which has been ‘better’?
Obviously, this is a subjective exercise, but here is my assessment.
This radar brings together the concept of ‘value’ via the On-Ball Value metric, with some other volume-related metrics included:
In some ways, the volume of shots a keeper faces are not something they directly control. For example, Dundee keepers have faced about triple the number of shots as Hart has, which creates far more opportunity to accumulate ‘value’ for a skilled shot-stopper.
We can see from the OBV radar that Gordon’s shot-stopping has been near the top of his peer group, while Hart has stopped roughly the number of shots Statsbomb’s model suggests he should have.
The combination of much higher volume in opportunities (shots faced), with better shot-stopping, and the importance in the weighting of shot stopping, results in Gordon having a wide lead in their Goalkeeper OBV metric. However, we can also see that Gordon’s Goalkeeper OBV percentile is below his shot-stopping. This is where the remainder of keeper play enters the picture- things like coming for crosses, shot handling and sweeping.
Here is a shot map showing only shots Gordon faced from crossed balls:
We can see from this report that Gordon conceded nine goals while facing 18 shots from crossed balls, with a Goal Saved Above Average (GSAA) metric of -0.14, so he saved about what would be expected from the quality of the shots he faced. However, look at the location of the shots! Perhaps unsurprisingly, in the Claims metric within the radars above, Gordon’s rank is in the 38th percentile.
Now we take a look at Joe Hart’s comparable shot map:
This has been a weakness in Hart’s performances at Celtic this season, and one I have highlighted before. He’s conceded eight goals from just twelve shots, with a GSAA of -1.56. The proximity of shots is not as close as Gordon’s, but the outcome is generally poor.
This is an example of how the relative degree to which a keeper ‘commands their box’ is separate from shot-stopping. Gordon has had a very good and generally consistent season of shot stopping, but at least some of the elevated number of shots he’s faced were likely directly attributable to his own shortcomings at commanding his box. The Goalkeeper OBV metric is one methodology to try and balance the different variables to assess overall keeper play.
As I scanned through keepers in the league this season while conducting this analysis comparing Hart with Gordon, a pretty clear ‘sleeper’ emerged. Jak Alnwick at St. Mirren has had a REALLY good season. Here is a scatter plot of keepers who have played at least 1,200 minutes and showing Goalkeeper OBV along with Shot Stopping percentage:
Alnwick has been a leader in both metrics, and his shot map facing crosses offers some insight as to the how and why:
He has conceded eight goals on 26 shots from crossed balls, with the four goals from inside his box having been scored by Liel Abada, Alfredo Morelos, Scott Wright, and Christian Ramirez, or the three opposing clubs with the highest wage bills in the league. St. Mirren have conceded about 1.35 npxG per 90 minutes this season, and Alnwick has faced 4.55 shots, or only slightly higher than Gordon’s 4.26.
Before anyone gets any ideas about Alnwick being a good candidate to play at a club like Celtic, his skillset has been a very good match for a club like St. Mirren given their style of play.
As the radar suggests, Alnwick has gone long a lot in Paisley and has not been asked to play out from the back much. He's been pretty poor at it when he has been.
For me Jak Alnwick has been both the most productive keeper in the league so far this season, and possibly the most valuable player at any club. With St. Mirren just six points above the relegation zone, the seven-plus goals Alnwick has prevented through shot-stopping, and likely even more with his all-around keeper play, is the difference between St. Mirren’s goal differential and Dundee’s. Gordon has had a good season at Hearts, and Hart a welcome stabilizing influence at Celtic, but, in my opinion, neither have approached Alnwick’s quality of play and importance to his club this season.
