I hate how the Golden Glove is awarded.

Basically, each league’s “Keeper of the Year Award” is given to the GK credited with the most clean sheets. Simply put, I do not think that clean sheets are the most important metric by which we should assess goalkeepers, as they can say more about the team’s defensive philosophy and the quality of the defenders in front of the keeper. Obviously, a good keeper contributes to holding the opposition scoreless, but if his team dominates possession, or sticks to an effective defensive tactic, he will face fewer shots and those that get through will be of lower quality.

Case in point: Martin Dubravka, filling in for the injured Nick Pope at Newcastle, has 5 clean sheets so far this season in just 10 starts. Emi Martinez has 3 in 28 starts. Whoever is in goal for Newcastle plays behind Sandro Tonali, Fabian Schar, and 6’6” of Dan Burn executing a physical, defensive gameplan. Martinez, on the other hand, is occasionally left wide open by failed counterattacks, and has most often worked with the central pairing of Pau Torres and Ezri Konsa, both of whom are better known for playing the ball than physical domination. I imagine most managers would take Emi over Dubravka, regardless.

I tried to come up with an alternative system, one based on player skill and the keepers’ overall game.

To my surprise, the numbers end up providing an endorsement of Ederson. This was not my intent. I didn't look for stats that would flatter him, nor did I work backwards to fit the numbers to him. I simply came up with my methodology after doing some reading and research, and it just so happens that Ederson is doing well this year in the areas I consider most impactful for a keeper. There is no fan-bias here.

Note: Stats are accurate up to Matchweek 29, League games only.

I looked at three areas, all of which I can attribute more directly to the keeper’s skill than the team’s qualities and the manager’s philosophy. Nothing happens in a vacuum of course; for example, passing to press-resistant, technically adept outfielders will increase a GK’s passing stats, regardless of his skill. But I wanted measurables that are less-dependant on factors outside the players control, and didn’t want to numb your senses with a wall of numbers. Thus, a three-pronged methodology and a way to apply it to several players at once.

For simplicity, I used the keepers with the most starts for each club. The 20 “Numbers Ones” of the Premier League…plus a composite “Average Keeper” - a theoretical player with statistics reflecting average performance of every keeper who has appeared in an EPL game this season. This statistical placeholder is what we measure real keepers against to determine their impact, positive or negative, on the team. In Fantasy Baseball terms, this is basically the generic “Replacement” player against whom we measure “Value Against Replacement.” I call him Joe Keeper. He provides what an EPL club should expect from a keeper, every game. I want to know who is better than him.

Here are the categories:

Personal Influence on Goals Allowed

Hardly a game goes by without a commentator or post-game analyst mentioning xG, or “eXpected Goals.” This fairly predictive measure of a team’s offense is made by comparing each shot on goal to thousands of similar attempts. If a shot of the particular type results in a goal 8% of the time, the attempt is awarded an xG of .08. Total the values awarded to each attempt and you have an idea of the offense a team could reasonably hope to have.

Example: an xG of 2.5 indicates a team should have scored twice or, with a little luck, gotten 3. It’s not perfect, of course; sometimes a team spams a bunch of low quality attempts, leading to a high xG, but few genuine scoring chances. Dumb luck and moments of great skill can also skew it, and that’s often when the pundits say someone is over-or-under performing xG.

Post-Shot expected Goals (PSxG) is a little more sophisticated, accounting for shot quality. “Basic” xG would award the same value to any two similar shots, for instance: twice, a striker receives a cross 15 yards from goal, in the center of the box, and shoots with his favored foot, aiming low-right. It’s a decent chance and is awarded an xG of .30. PSxG accounts for a scenario where one of these shots is mis-hit and dribbles to the keeper but the second is a thunderbolt that no one could be reasonably expected to save. The values change to .05 and .75 respectively, marking one as far a more “quality” chance.

The lunatics at Fbref.com not only track this stat for shooters, but also keep a tally of how “good” the shots each keeper faces, their accumulated PSxG-faced. They compare this number to the number of actual goals conceded, resulting in what I think is the truest measure of how good a shot-stopper each player is: a plus/minus comparison to what he “probably” should have allowed. After all, every keeper will face shots and it will sometimes come down to how good he is at getting to the ball.

I further looked at “Mistakes Leading to Goals” - the regrettable moments when a player makes a mistake bad enough that the official scorer notes his culpability in the scoresheet. I subtracted such mistakes from the PSxG, creating a rating that tells us how often the keeper personally saved a goal - or failed to prevent the team from conceding. A high score indicates either consistent skill at shot stopping (think of someone who seems 10 percent more likely to stop a good shot than the average GK) or has that undefinable last-chance reflex, miraculously turning away a few seemingly-certain goals each season. A weaker keeper will have a negative score, simply allowing goals on shots a player at this level is reasonably expected to get to, or making glaring mistakes.

The theoretical “average keeper” will have a score of 0.0 or close to it, not making obvious mistakes and saving the shots that he should, but rarely making great saves as the last line of defense.

Here is how the boys did:

Presented visually, left is mistakes and “he should’ve had that” - right is consistency and/or big saves

Vicario is in a class by himself, but Ederson scores well, and was second before factoring in his two MLGs.

Launched Goal Kick Accuracy

Every keeper is expected to boot it on occasion, but some teams are more committed to playing it out from the back, or favor middle-distance attempts over long-balls. Considering this, I looked at completion percentages on “launches” rather than frequency and relative depth of goal kicks. A “launch” is defined as a goal kick driven downfield, travelling 40 yards or more in the air, then received by a teammate. A keeper who does this well can start attacks from nothing, even garnering the occasional key-pass or assist. How often those chances arise has a lot to do with how the manager sets up. Vicario, for example, has attempted a paltry 28 versus Jordan Pickford’s league-high 598 and Mat Sels’s 513. This disparity is all on the tactics, but how accurately the keeper launches is (mostly) down to his skill and familiarity with his teammates. Thus, everybody’s Launched GK success percentage.

Eddy and Emiliano are in a class by themselves, though Pickford’s and Sel’s respectable accuracy intersect nicely with their respective teams’ love of the long-ball to create something dangerous.

Claim Authority

Crosses are fundamental - one of the main methods of chance creation for many squads. Defenses have to deal with them. Good wide defenders will reduce the frequency and quality of crosses getting into the penalty area, and powerful and/or savvy centerbacks will deflect or intercept their share. But a few will get through, and some of them into that dangerous space on the six-yard line where it is difficult to determine who should handle the situation. Considering this, I looked at claim percentage rather than total crosses claimed.

Claims - defined here as a lofted ball pulled down by the keeper, then passed or thrown to an outfielder - are part of the job for a modern keeper. Ripping down a ball instantly stops an offensive move and potentially starts a counter.

Tactics and personnel will influence how many crosses a keeper faces, so I looked at what share of crosses a keeper claims out of the total his team faces, rather than total claims. I’ve ranked our starters by the percentage of all crosses that enter their area which they “claim.”

Joe Keeper gets to 7.1% of lofted balls in his penalty area, so our top guys really change the look of a defense, with Muric and Sanchez succeeding twice that often. Alisson gets to half as many, though I suspect this boils down to Van Dijk handling this chore better than most defenders.

So, my evaluation can be summarized with a three-part stat-line. PSxG/MLG, launch accuracy, and cross claim percentage. For Joe Keeper: 0.1/33.5%/7.1%. We can assess a keeper’s impact on the game, positive or negative, by looking at where his numbers deviate from this expected norm. This chart visualizes the idea (poorly) - basically, just follow the lines from the top bars to see how far above or below expectations each player performs.

As you can see, exceeding the mean in all three areas is a tall order. Each player has strengths and weaknesses. So, I ranked the boys in all three categories, and then averaged their rankings - creating a “rank of ranks.” But that is a stupid name, so I called it Impact Rank - and I feel it is a composite rating of the most impactful things a keeper can do, and how well he does them, relative to his peers.

Eddie is first, to my surprise, narrowly edging out Flekken, who I also did not expect to score so well. They are joined by Raya, Martinez, Vicario, and Kepa as the few starters more impactful than our theoretical Joe. Muric, Hermansen, and Pope are also right there. Jose Sa is uniquely underperforming this year, with only a decent claim percentage improving his dismissal score.

In closing, I don't intend this as an overall ranking of the keepers. This was simply an exercise in attempting to measure the impact of each player's skill-set on the game. Eddy, for his faults, is evidently a game-changer. Several other keepers, despite their reputations, do not reach the standard of our man Joe, the totally average pro.