Expected Saves

Understanding expected saves

Expected saves (xS) has become an essential tool in goalkeeper analytics, offering deeper insights than traditional metrics like save percentage or clean sheets. It evaluates a goalkeeper’s performance by assessing the probability of stopping specific shots based on context and historical data.

What are expected saves?

Expected saves quantifies the likelihood of a goalkeeper making a save for a given shot, taking into account shot characteristics such as distance, angle, shot type, body part used, and goalkeeper positioning. Instead of simply counting saves and goals conceded, xS provides a probability-based assessment of how difficult each save attempt was.

Relationship with expected goals (xG)

While xG calculates the chance of a shot becoming a goal, xS represents the likelihood of the same shot being saved. In basic models, xS is derived as 1 – xG, but this can be overly simplistic. More sophisticated models factor in additional variables like defensive pressure, reaction time, and goalkeeper reach to produce more accurate and detailed evaluations.

Importance of expected saves

Integrating xS into analysis enables clubs and analysts to:

– Evaluate goalkeeper performance: Determine how a goalkeeper performs relative to the difficulty of the shots faced, helping to identify those who consistently exceed expectations.
– Inform scouting and recruitment: Spot underrated goalkeepers who excel in high-xG situations, which may be missed by traditional stats.
– Enhance training programmes: Use xS data to identify patterns in the types of shots a keeper struggles with and design drills to address those specific weaknesses.

Methodology: Calculating expected saves

Developing a robust Expected Saves (xS) model requires detailed shot-level data and advanced statistical modelling. Below is a breakdown of the essential components and methodology used in constructing an effective xS system:

Data collection and preparation

Accurate xS calculations begin with collecting rich, granular data points that influence whether a shot is saved or results in a goal. Key attributes include:

– Shot location: The pitch coordinates from where the shot is taken.
– Shot angle: The angle relative to the goal, impacting how much goal area the keeper must cover.
– Shot distance: How far the shot is taken from the goal.
– Shot type: Classification such as header, volley, or ground shot.
– Shot velocity: The speed of the ball when struck.
– Goalkeeper positioning: The keeper’s location at the time of the shot attempt.

Modeling approach

XS models rely on statistical methods suited for binary classification (goal vs. save). Common approaches include:

– Logistic regression: Estimates the probability of a save based on shot characteristics. It’s easy to interpret and widely used in early models.
– Machine learning algorithms: Techniques like Random Forests, Gradient Boosting (e.g., XGBoost), and Neural Networks offer greater predictive power by modelling non-linear relationships and interactions between variables.

Key variables in the model

Several inputs strongly influence the probability of a save:

– Distance to goal: Closer shots leave less reaction time, lowering save chances.
– Shot angle: Narrower angles favour goalkeepers; wider angles require more coverage.
– Shot velocity and trajectory: Faster or dipping shots are harder to save.
– Goalkeeper positioning: Being well-positioned increases the likelihood of a successful save.
– Defensive pressure: May obstruct the goalkeeper’s view or reduce the quality of the shot.

Model validation and performance evaluation

To ensure the xS model is reliable, performance must be validated using various metrics:

– Accuracy: The proportion of correct predictions the model makes (both saves and goals).
– Precision and recall: Evaluate how well the model identifies actual saves (precision) and captures all true saves (recall).
– ROC curve and AUC (Area Under Curve): Assess how effectively the model differentiates between saves and goals at various thresholds.

Applications of expected saves

The expected saves (xS) metric has become an invaluable tool in modern football analytics, offering deeper insights into goalkeeper performance beyond traditional statistics. By quantifying the probability of a save based on shot characteristics, xS facilitates more nuanced evaluations and strategic decisions across various domains.

Performance valuation

In terms of performance valuation, xS fills a crucial gap left by metrics like save percentage and clean sheets, which often overlook the quality of shots faced. Instead of simply counting saves, xS assesses the difficulty of each save attempt by considering variables such as shot distance, angle, and speed. For example, a goalkeeper who regularly saves high xS shots demonstrates notable technical ability, even if their overall save percentage isn’t exceptional. This level of granularity enables coaches and analysts to better assess performance, identify underappreciated strengths, and detect subtle areas for improvement that conventional stats might miss.

Scouting and recruitment

When it comes to scouting and recruitment, xS helps clubs uncover talent that may be flying under the radar. By highlighting goalkeepers who outperform expected benchmarks, teams can identify individuals who show consistent shot-stopping excellence, particularly against high-value chances. These insights offer an edge in transfer decisions by quantifying skills that may not be apparent through visual scouting alone, making xS a key tool for spotting undervalued or overlooked talent in the market.

Training and development

From a training and development standpoint, xS is instrumental in shaping goalkeeper training programmes. By isolating specific shot types or scenarios where a goalkeeper underperforms such as long-range efforts or close-range one-on-ones, coaches can create focused drills to address those areas. Monitoring xS trends over time also provides a performance feedback loop, enabling staff to measure improvements and adjust training plans accordingly..

Tactical decision-making

On a tactical level, xS serves a broader purpose by informing defensive strategy. Teams can use it to evaluate how well their tactical setup protects the goal. If the data shows a pattern of conceding high xS opportunities, it may indicate lapses in defensive structure or pressing inefficiencies. Addressing these systemic issues not only supports the goalkeeper but also strengthens the team’s overall defensive resilience.

Case studies: Real-world applications of expected saves

– José Sá – Wolverhampton Wanderers (2023/24 Premier League season): Wolverhampton’s Jose Sa faced 66.45 expected goals (xG) based on shots on target and conceded only 57 goals.
xS impact: Outperformed expectations by 9.45 goals, showcasing elite shot-stopping ability and critical importance to Wolves’ defensive strength.

– David De Gea – Fiorentina (2023/24 Serie A season): Achieved a 73% save percentage, the highest in Europe that season.
xS impact: Reflected his adaptability and proficiency against close-range shots, highlighting how xS reveals the effectiveness of a comeback season.

– Dean Henderson – Sheffield United (2019/20 Premier League season): Faced 39.4 expected goals on target (xGOT) and conceded only 32 goals.
xS impact: Prevented over seven goals above average, a key contributor to Sheffield United’s strong defensive record and a standout xS performer.

– Thibaut Courtois – Real Madrid (2021/22 UEFA Champions League final): Made nine saves in the final, the most in a UCL final.
xS impact: Critical in securing a 1–0 win against Liverpool, demonstrating how xS can highlight decisive performances in high-stakes matches.

Limitations and considerations

Expected saves (xS) is a great tool for analysing goalkeepers. But to use it well, we need to know its limits and what to consider.

– Data quality and granularity: The reliability of xS models relies heavily on detailed, accurate data. Variables such as shot location, speed, and goalkeeper positioning must be precisely recorded. Poor-quality or incomplete data can distort performance analysis.
– Model assumptions and variability: Different models use different methodologies, some simply subtract xG from 1, while others apply more advanced logic. This inconsistency can make it difficult to compare xS scores across platforms or studies without context.
– Contextual factors: Many xS models don’t fully account for real-world conditions like defensive pressure, weather, or match intensity. These elements can greatly affect save probability but are difficult to capture in standard datasets.
– Sample size and variance: Because saves and goals are relatively low-frequency events, short-term xS data can be volatile. A few unusual games can skew interpretation, especially over small sample sizes.
– Interpretability and communication: xS is a statistical measure that may not be immediately clear to coaches, players, or fans used to traditional stats. To make it actionable, results need to be clearly visualised and explained.

Sportmonks and expected statistics

While expected saves (xS) is a valuable metric for assessing goalkeeper performance, its effectiveness hinges on the availability of detailed and accurate data. Sportmonks addresses this need by providing comprehensive football data through its football API, supporting a wide range of statistics essential for advanced analytics.

Sportmonks football API

Sportmonks’ football API offers extensive coverage, including

– Shot data: Information on shot location, type, and outcome.
– Goalkeeper statistics: Metrics such as saves, goals conceded, and clean sheets.
– Expected goals (xG): Data on the probability of a shot resulting in a goal, available across various packages.

These data points are crucial for constructing robust xS models, enabling analysts and developers to derive meaningful insights into goalkeeper performance.

Building expected saves models with Sportmonks data

Although Sportmonks does not currently provide a dedicated xS metric, the rich dataset available through its API allows for the development of custom xS models. By leveraging shot characteristics and goalkeeper statistics, analysts can estimate the probability of saves, facilitating a deeper understanding of goalkeeping effectiveness.

Flexible access and integration

Sportmonks offers various subscription plans, including a free trial, to accommodate different needs. The API is designed for easy integration, with comprehensive documentation and support to assist users in incorporating data into their applications or analyses.

For more information on Sportmonks’ football API and its capabilities, visit their official website.

Build smarter goalkeeper insights with real match data

Expected Saves (xS) reveals how well goalkeepers truly perform, but you need the right data to make it work. With Sportmonks’ football API, you can access detailed shot locations, xG values, and keeper stats to build your own xS models or enrich your football app.

Track key shot and save metrics, compare performances across leagues, and spot trends that traditional stats miss.

Start your free trial with Sportmonks today and bring deeper goalkeeper analytics to life.