Summary of the Key Points
This article reveals how the data we often see in football matches—such as distances covered, expected xG (expected goals), and player ratings—are generated, as well as the commercial logic and limitations behind them. In simple terms, football data consists of two types: “event data” recorded manually (who passes to whom, who makes a tackle, etc.) and “tracking data” captured by machines (distances covered, positions, etc.). This raw data is then processed into actionable indicators like xG and player ratings. Data companies make money by selling this information to media outlets, football teams, betting companies, and AI platforms. However, there are limitations to these datasets; they cannot fully understand tactical intentions or the human elements that make football unique.
Where Does Football Data Come From?
Football data is collected in two ways:
1. Event Data: Recorded manually to document what happens during the game. For example, who passes the ball at which minute, who makes a tackle, and who takes a shot. This is typically done by teams of three people: two monitor each team in real-time, and the third person reviews the footage to verify the records. When recording a pass, data collectors use a mouse to mark the starting and ending points, and the system identifies the passer.
2. Tracking Data: Gathered by machines to track the movements of players and the ball. High-speed cameras are installed around the field, using computer vision technology to capture numerous frames per second and determine the precise positions of players and the ball in real-time. Players wear special “sports underwear” with sensors that provide objective physical data such as position, speed, and heart rate.
xG and Player Ratings: Not Random, but Calculated Based on Historical Data
The raw data serves as a foundation for generating meaningful indicators:
- xG (Expected Goals): This measures the likelihood of a shot converting into a goal. For instance, a penalty has an xG of around 0.76 (meaning 2-3 out of 10 attempts will result in a goal), while a long-range shot from outside the penalty area might have an xG of only 0.03 (only 3 goals out of 100 attempts). Models are trained using data from hundreds of thousands of past shots, considering various factors like shooting distance, angle, whether it’s a header or a footshot, the number of defenders, and the goalkeeper’s position.
- Player Ratings: These ratings are calculated by weighting all the player’s actions throughout the game based on their role (e.g., passing and shooting in the penalty area). For example, Sofascore gives Messi a score of 10 and Ronaldo a score of 6.1. The rating system takes into account the context of each action and adjusts dynamically (almost 2,000 evaluations per match), with winning players generally receiving higher scores. However, ratings focus solely on actions and ignore tactics; for instance, a defender who focuses solely on guarding Haaland might have a low rating even if they perform no offensive tasks.
How Do Data Companies Make Money?
Developing and maintaining a data system is costly, but data companies profit significantly by selling it to various clients:
1. To Media: They provide real-time statistics for broadcasts (e.g., “So-and-so made 26 touches during the match”) and analysis for commentators (e.g., “He has the most tackles in the league”). Opta serves over 800 clients annually, including FIFA and UEFA.
2. To Football Teams: Teams use data to select players and devise tactics (e.g., identifying fast, accurate midfielders or studying opponents’ weaknesses). Training data (such as heart rate and distance covered) helps coaches identify players who need improvement.
3. To Betting Companies: These companies use data to adjust betting odds in real-time, minimizing risks and generating substantial profits.
4. To AI Platforms: FIFA’s Football AI Pro platform uses data for tactical analysis, giving weaker teams access to information on stronger opponents, which has contributed to many upsets in World Cups.
The Limitations of Data
Despite the detailed nature of football data, there are still gaps:
- Tactical Understanding: Data cannot fully capture strategic decisions or the subtle dynamics of gameplay, such as a defender’s dedicated focus on Haaland.
- Human Elements: Moments of teamwork and spontaneous creativity (like Messi’s “Hand of God” goals) are beyond the scope of data analysis.
- Ratings vs. Performance: Experienced fans often trust their own observations over ratings, as they consider the context behind each action.
In summary, football data is a valuable tool for understanding matches, but it cannot replace the emotional and unpredictable aspects that make football so captivating.
(The entire article is written in plain language to make it accessible to non-financial and non-sports enthusiasts.)