Reassessing Goalkeepers (Again): How Much Do They Influence Shooting?

Reassessing Goalkeepers (Again): How Much Do They Influence/Affect Shooting?


Goalkeeper metrics and advanced data are still far from perfect, that’s not entirely news. It wasn’t until a few years ago that basic metrics such as save percentage were the barometer for what determined good goalkeeping – Save a lot of shots, and voila! That’s a good goalkeeper. However, with a lot of basic aggregation statistics that provides tons and tons of issues for evaluating players. For example: Mark Flekken of Freiburg, currently has a higher save percentage than Manuel Neuer. While the Freiburg stopper is a good player, are we led to believe that Neuer is worse than he is? Even metrics like goals conceded per 90 are conditioned by tons of external factors (team strength for one) and are not indicative of individual quality. 

SInce these days, we’ve now moved onto more advanced metrics/models which help better grasp how goalkeepers perform: Post shot expected goals, goals saved above average, etc. are just some basic examples of these. They help contextualize more information on the goalkeeper (their position, habits in goal) as well as the information of the demands placed on them by the shot (shot height, end location of the shot, etc.) rather than simple “accumulation” statistics. With all this information in our grasp, and readily available for analysis, how do we link this higher level of data/analytics with traditional coaching habits and ideas on what constitutes good goalkeeping?

An example of how increased stability/sophistication surrounding goalkeeper metrics & data has given us newfound insight into how they rate within their position.

Growing up as a goalkeeper myself in the American college system, training with MLS/USL clubs as well as my brief coaching of goalkeepers before moving into performance analysis, I was always taught and taught myself about ways (particularly in 1v1s) about how the player can “make the opposition miss” Whether it be making sure your angles are spot on to take up as much of the goal as possible, to pretty much the fugazi idea of having an imposing presence to make opponents lose composure. While much has been done in improving how we evaluate goalkeepers in terms of saving shots, not much has been done in regards to putting objective findings behind how they affect misses. Simply put – Can good goalkeepers influence how often the opposition hit the target? That’s my end goal, and I think I’ve found some interesting findings thus far!

Before we dive into it, the data set: I’m using goalkeeper data (and by extension shooting data) going back to the beginning of 2018 up until the start of this season from the “Big 5” leagues in European football. While most of the data has been unedited, I’ve manipulated some of the data myself to help better provide situational context to the players as needed – This will become clear as you read on.

Are Shots Misses Random?

Obviously, the question of this article is meant to discover what determines shot misses – Because of this, it’s implied that shooting % is not random and that they are repeatable phenomen. In other words, some kind of external factor (such as the goalkeeper) has an effect. This would mean that there is some correlation between shooting percentage  season after season. 

As we can see from the image below, which details the shooting % (i.e. judging accuracy and shot misses of various positions within the dataset), we can see that to be emphatically true: Around 30% of shots from players are on target. Essentially, a player’s ability to hit the target and create/score chances is consistent and repeatable as suspected. This is good to know, considering the entire root idea of one of football’s fundamental analytics concepts, expected goals, is reliant on this sort of recurrent phenomena. However, we have not nailed down *what* makes this so: Who or what is the root cause of shooting percentages being consistent (or inconsistent for certain players!)

How Do Goalkeepers Affect Attackers & Shots?

Since we’ve now found out that missed shots are: A) repeatable, and B) caused by some outside factors, we need to take a look at what makes them so:

  • Is it attackers simply being bad shooters season after season?
  • Is it the defenders or the opposition? (Good structures forcing bad shots, blocking shots, denying the attackers ability to score, etc.)
  • The goalkeeper: Their habits in goal, be it positioning, ability in certain scenarios to affect/negate shots.

Looking at the goalkeeper specifically requires a bit more granular detail, but one thing we do know is that since every goalkeeper (assuming they play a decent threshold of minutes) plays against different attackers and teams, we can isolate them out of context. The dataset used to analyze these habits is also broad enough, and features enough filtering that we are not reliant on small sample sizes. The other reasons for shot misses requires more of a deep dive into external factors. 

One way to do this is to look at various positional groups to see how their shots/shot types, etc. compare. The results are not especially surprising: Players further away from goal (defenders, midfielders, etc.) are less likely to hit the target compared to attack minded players – Hence why they play where they do. This is also a consequence of the scenarios they find themselves in front of goal. Center backs, full backs, central midfielders will usually be further away from goal or getting shots off on set pieces: By their very nature occasions where it’s tougher to hit the target. Attacking players are the beneficiaries and finishers of sequences to goal, and create a high volume of them. This isn’t news, but it’s good to see it visualized. 

In the dataset, the table below on the left shows the goalkeepers (within a decent threshold of minutes played) that have the lowest shooting % against. For reference, the average sits around 33% of all shots faced. The table on the right lists the top ten goalkeepers who face the most. Can we find any real connection between the two groups? Not really: Both sides of the list have keepers who are rated highly as well as those who are relative unknowns. It’s important to note, of course, that team style for this type of data cannot be ignored.

  • Do they primarily sit in a low block (leading to shots from range, higher shots per 90 against, etc.) 
  • Does the team like to possess the ball, and might give up less chances, but chances which might be higher xG (I’m looking at you Bayern Munich and Maneul Neuer).

Obviously, these caveats require further examination, especially related to the style of the individual goalkeeper himself. 

Goalkeeping Style/Traits

Every goalkeeper is not the same, and these “profiles” of goalkeepers are important to provide context for how they affect play. If we see the graph below, without considering these types of individual playing styles, the data shows us that there is very little correlation between how many shots on target (goals + saves) a goalkeeper deals with, and how they rate amongst their peers. Goals Saved Above Average, a metric I adopted for this data set from StatsBomb (which you can read about here) essentially takes a look at how many expected goals they should concede, alongside how many they actually concede – Those who rank highly concede less. As the graph displays there is a slight inverse correlation, but it is too small for us to concede that: Better GKs = Less shots on target. As mentioned above though, what about various goalkeeping types? I’m going to group players into two broad categories – I recognize this isn’t 100% optimal: There are more than two types of keeper,  but in the early stages of analysis it helps serve it’s desired purpose. 

The first of these is keepers I’ve classified as “line goalkeepers” – Those who are less active with how they position themselves in the goal, less likely to sweep off their line, and more likely to let their defenders deal with anything in the box. This grouping/bin of players was achieved by mixing together their starting position x and y coordinates, as well as how often they moved off their line to preemptively play the ball. It’s not a perfect grouping, but it essentially achieves the same goals. 

Compared to the entire dataset of players, there isn’t much difference: A slight inverse correlation (i.e. better GSAA leads to slightly fewer shots faced on target) but it’s really not anything to write home about – There are some huge variations here. 

The second grouping is the “proactive goalkeeper” – A goalkeeper who is more likely to come off their line and prevent danger before it occurs (i.e. shots) rather than let it happen, or generally just performs more actions higher in relation to the goal-mouth than their peers. The flip towards a positive correlation is massive and has some important ramifications on what it means for assessing a goalkeeper’s effect on how often opponents hit the frame. Essentially, the goalkeepers who are more active in regards to their playing style give up more shots on target – What does this all mean? It’s complicated.

Furthering Separating Out Goalkeeping Profiles

Can we see any trends within these two profiles/bins to try and figure out why proactive keepers face more shots on target? Let’s see. Obviously, two groupings here for goalkeeper profiles are not exhaustive, nor are they especially advanced in their qualifications. Any fan of football could tell that within line goalkeepers there are those who are deeper in their set position, but often negate a lot of potential shots by coming for crosses (i.e. lessening the potential shots they face by coming to claim)  – The data is inconclusive on this. As shown below, within the bin of players highlighted certain keepers face more shots on their goal than their peers across the board – There is no correlation that coming for crosses alleviates this. 

For the proactive goalkeepers, this sort of aggressive behaviour manifests itself in a manner of ways: Not everyone is going to play like Neuer, Alisson, Ederson, and act almost as an auxiliary center back – This nature of their game can also just be in how they like to play off their line in facing shots: Their height in relation to the goal-line and the shooter where they set. Below are two examples – One of a higher set position, and another more conservative version. It’s important to note that there is not right or wrong with these: Every coach knows that it’s important to allow the goalkeeper to play how they are comfortable (within reason). The issues with these come more so if they are set suboptimally laterally rather than horizontally when facing shots.

Ederson displaying two contrasting approaches to positioning off his line when dealing with attempts on goal.

Much like the other band of data listed above, there is no real true correlation between higher set positions and conceding more shots – However, the variance across the board in this goalkeeper set is much more wild and split than the reactive goalkeepers: Something is at play here within this sub category that is leading to them making more saves. We can immediately rule out team strength or individual player quality: Alisson, Bürki, Ederson, Neuer, etc. are all listed above, and no one would consider them to be substandard. Something else is at play here, but it helps give an interesting thread to follow.

What Does This All Mean?

While football is obviously an incredibly fluid game, nothing happens in a vacuum, and the goalkeeper is the player who is most affected by the opposition (and by extension, his teammate’s reaction) to the ball and where they are able to generate chances: Even keepers who snuff out danger before it happens are going to only have a small effect on how often they are worked. It’s important to touch on these contextual circumstances:

For one, the basic team strength has a big factor on shots they face on target. If a player in goal for a team at the bottom of the table, with a terrible xGA record is facing 45% of all shots against on target, we can realistically say they will face more shots than a team competing at the top of the league. Use the two charts below as an example. These are two shots a player faced over a 5 game span. Who would you rather be? Certainly the individual on the right. Over the course of an entire season, these shots add up: More shots + better locations (for the attacking team) = more shots on target. Pretty simple right? 

Two comparisons of shots goalkeepers faced – One in a strong team, and the other in a weaker side.

However, a spanner is thrown in the works when we remember that within the two groups profiled, there are goalkeepers across the spectrum in regards to the strength of the club they play for. This forces us to look at team style, and how one’s game model affects it.

Recently, at the StatsBomb Conference just over a week ago now, John Harrison & Max Odenheimer of LAFC spoke about how a player’s style fits within a team’s style (and by extension how to train and recruit for that player). Without spoiling the main aspects of their presentation, the two asserted that teams should sign goalkeepers who thrive in certain scenarios/situations that their team often gives up chances from. For example: Liverpool have Alisson who is one of the best 1v1 players you’re likely to see. Liverpool, despite dominating the majority of the ball, often are prone to one of these big chances every few games. They are comfortable with this though – It’s one of the tradeoffs they are prepared to make. Guardiola’s teams in terms of how the goalkeeper plays in possession are another similar example. What does this mean?

Alisson, of Liverpool, dealing with a 1v1 opportunity: A type of save they concede as a trade-off within their system of play.

It is essentially the inverse of the player who plays for a bad team: Rather than giving up higher shots on target due to accumulation and poor defending, the Alissons and Edersons of the world are forced to make less saves, but they often are from high xG locations – Thus making shots on target much less of a difficult thing for attackers (closer to goal, less bodies with an ability to block it, etc.) This a prime example of an individual trait/playing style of the goalkeeper meshing with his collective unit. 

I present this contextual information to ram home one point: The early, soft-analysis of shots faced on target leads me to believe that something about more proactive goalkeepers are more likely to be called into action, ever so slightly, compared to their peers. 

Going Forward

As you can see, this is a complex conundrum within the data (and by extension how we coach/analyze/view the position itself) to answer. It requires a much more thorough understanding and dive into the numbers and what creates them: Something I intend to do. However, based on an early dive into the data there seems to be a tentative link between more proactive styles of goalkeeping  and shots they are forced to save. As I’ve noted throughout this is not a call for all keepers to sit on their line and adopt a traditional style of play that was more common 20+ years ago. Rather, it’s my goal to see what causes this imbalance, both individually and collectively. What is needed on my end? Metrics on the position have come a long way recently, as well as how we understand it – However, the context around how and why proactive goalkeepers act the way they do needs to be applied: Especially what scenarios they are facing shots from as opposed to just raw data. This series going further will utilize more video to show in-match, examples of proactive/reactive styles and to help get a firmer grasp on the initial question: How much do goalkeepers influence shooting?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: