Skip to main content
Defense

GK Shot-Stopping in Football

Goals Saved Above Expected (GSAx) measures how many goals a goalkeeper has prevented compared to the average goalkeeper facing the same shots. A positive GSAx means the keeper is performing better than expected; negative means worse.

Defense
Category
Current season only
Season Scope

In a Nutshell

Think of a goalkeeper as the last line of defence. Some keepers stop shots that most goalkeepers would let in — they are "saving above expectation." Others let in shots that most keepers would save. This signal measures exactly that gap: how many goals has the keeper prevented (or failed to prevent) compared to what an average keeper would do facing the same shots? A positive number means the keeper is outperforming; negative means underperforming.

Why It Matters

Elite shot-stopping can be the difference between winning and losing tight matches. A goalkeeper in excellent form can provide a defensive foundation that allows the team to take more attacking risks. Conversely, a struggling keeper can undermine even strong defensive play.

How It's Calculated

Data Sources

Post-shot Expected Goals (PSxG) data from match analytics providers. PSxG measures the quality of each shot after it is taken, accounting for placement, power, and angle. Combined with actual goals conceded to compute Goals Saved Above Expected (GSAx).

Key Formula

Measures how many goals each keeper saves (or fails to save) compared to what an average keeper would do facing the same shots. With limited playing time, the system is cautious and assumes the keeper is closer to average. Only after a substantial number of matches does the keeper’s actual record carry full weight.

Normalization

The difference between the two goalkeepers’ shot-stopping records is compared against the realistic range of goalkeeper quality, so the system can distinguish a genuinely elite or poor keeper from one who is simply average.

Confidence Impact

Both goalkeepers must have data for a meaningful signal. If both are missing, uncertainty is 0.95. If one is missing, a +0.40 penalty applies. Below 900 minutes (~10 matches) per keeper, a +0.15 penalty is added. Starter certainty below 70% adds another +0.15 — if it is unclear who will play in goal, the signal is less reliable.

Example

See how this signal played out in actual matches.

Liverpool vs Tottenham

Premier League — Dec 2024

Liverpool
Strong advantage
Signal leaned clearly toward home side
Tottenham
Disadvantage
Metrics reflected weaker position

What happened: Liverpool's metrics were consistently superior across this signal, aligning with the eventual 6-3 result. The signal correctly identified a significant imbalance.

View this match on xGoal

Man City vs Everton

Premier League — Dec 2024

Man City
Neutral
No clear advantage from this signal
Everton
Neutral
Similar metrics on both sides

What happened: Both sides showed similar metrics for this signal — neither had a meaningful edge. The match ended 1-1, consistent with the balanced assessment.

View this match on xGoal

Aston Villa vs Arsenal

Premier League — Dec 2025

Aston Villa
Slight edge
Signal indicated moderate home advantage
Arsenal
Competitive
Close but slightly behind on this metric

What happened: The signal leaned slightly toward Villa, but Arsenal's strengths in other signals offset this. This illustrates why no single signal tells the full story — context from all signals matters.

View this match on xGoal

Common Misconceptions

Myth: "A keeper who concedes few goals is automatically good"

Reality: Goals conceded depends on the defence in front of the keeper. A goalkeeper behind an elite defence may face only easy saves. GSAx isolates shot-stopping ability by comparing against what an average keeper would do facing the same shots.

Myth: "Three matches is enough to judge a goalkeeper"

Reality: With only 270 minutes, the Bayesian shrinkage factor is just 0.23 — meaning 77% of the signal comes from the league average, not the keeper’s actual performance. It takes roughly 20 matches (1,800 minutes) for the data to carry two-thirds of the weight.

Myth: "The better goalkeeper always wins the match"

Reality: Shot-stopping is one factor among many. A team can have the better keeper but lose because of poor chance creation, defensive errors, or set-piece vulnerability. This signal measures goalkeeping quality in isolation — it does not predict outcomes on its own.

Advanced Insights

What This Signal Cannot Tell You

  • No single signal determines match outcome
  • All signals carry uncertainty that widens probability ranges
  • Data freshness affects confidence in the assessment
  • Context matters: signals interact with each other and with match circumstances

The Bottom Line

Goals Saved Above Expected (GSAx) measures how many goals a goalkeeper has prevented compared to the average goalkeeper facing the same shots. Use it alongside other signals for a balanced view of the match.

Related Signals

This signal works alongside others to build a complete picture.

Key Terms

Understand the terminology used in this signal.

Further Reading