Temporal dynamics of goal scoring in soccer
Guteraa Ayana, Alexander Ehlert ∗, Joseph Ehlert, Luca Santagata, Maddalena Torricelli, Brennan Klein · 2025
Read Temporal dynamics of goal scoring in soccer by Guteraa Ayana, Alexander Ehlert ∗, Joseph Ehlert, Luca Santagata, Maddalena Torricelli, Brennan Klein (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-18606.txt.
Introduction
Guteraa Ayana, Alexander Ehlert, Joseph Ehlert, Luca Santagata, Maddalena Torricelli, and Brennan Klein wrote a 2025 paper on when goals are scored in soccer matches. Goals are often treated as random events spread evenly over time. The authors test that assumption with event data from 3,433 matches across 21 leagues and competitions.
Goals proved more likely as matches wear on, and the early minutes of each half saw fewer goals than chance would suggest. Gaps between goals shrink in an exponential pattern, so goals cluster in time. Splitting by which team scores next showed a burst effect: the team that just scored is more likely to score again soon. Fatigue, tactics, and momentum are candidate drivers. The findings can sharpen live analytics and coaching decisions.
Catalog record
| CREATOR | Guteraa Ayana, Alexander Ehlert ∗, Joseph Ehlert, Luca Santagata, Maddalena Torricelli, Brennan Klein |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-18606
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Temporal dynamics of goal scoring in soccer Guteraa Ayana, Alexander Ehlert ∗, Joseph Ehlert, Luca Santagata, Maddalena Torricelli, Brennan Klein Abstract We investigated the temporal distribution of goals in soccer using event-level data from 3,433 matches across 21 leagues and competitions. Contrary to the prevailing notion of randomness, we found that the probability of a goal being scored is higher as matches progress, and we observed fewer-than-expected goals in the early minutes of each half. Further analysis of the time between subsequent goals shows an exponential decay, indicating that most goals naturally cluster closer together in time. By splitting this distribution by the team that scores the next goal, we observe bursty goal-scoring dynamics, wherein the same team is more likely to score again shortly after its previous goal. These findings highlight the importance of match context—whether driven by fatigue, tactical adaptations, or psychological momentum—in shaping when teams are able to score. Moreover, the results open avenues for extending data-driven methods for identifying high-impact moments in a match and refining strategic decision-making in soccer’s evolving analytical landscape. 1 Introduction Soccer, also known as football throughout most of the world, is a sport played between two teams of 11 players on a rectangular field with a goal at each end. Soccer is governed by the rules of the game known as the Laws of the Game, established by the International Football Association Board [ 1 ] . The objective of the game is to score goals by maneuvering a ball into the opposing team’s goal using any part of the body except the hands and arms, with the exception of goalkeepers who are allowed to use their hands within a designated box. Matches are divided into two halves of 45 minutes each, with a brief halftime interval. The half length can be extended to account for stoppage time coming from player injuries, substitutions, and more. With its simplicity and minimal equipment requirements, soccer has grown into the most widely played and watched sport in the world [ 2 ] . Analytics plays an increasingly prominent role in contemporary professional soccer, influencing team tactics, player contracts, and coaching strategies. Before this popularization of data-driven methodologies, soccer tactics were predominantly shaped by intuition, experience, and observational analysis [ 3 , 4 ] . Statistical modeling and machine learning algorithms are now widely employed to analyze player performance, predict match outcomes, and optimize game plans [ 5 ] . Today, the average television viewer of professional soccer encounters substantially more statistical information, with metrics like the Expected Goals (xG), pass success rates, and heat maps now commonly used to summarize key patterns in a game. This growing understanding of the role of data in professional soccer has coincided with a new range of tactical innovations and trends. Amid this heterogeneity, we also see a renewed interest in uncovering new ways of quantifying and characterizing commonalities between teams. In this work, we document gameplay trends that persist across countries, leagues, and seasons. Specifically, we measure the time that goals are scored in games and—in games with two or more goals—the time difference between subsequent goals. Drawing on a large dataset of professional matches across multiple leagues and competitions, our analyses focus on identifying distinct moments within games that may elevate or reduce the likelihood of scoring, relative to a null model of random goal timing. By contrasting observed events against this theoretical framework, we find when teams are most/least likely to score a goal. These observations motivate a deeper look into whether late-match attacking pushes, early-half lulls, or other situational shifts can decisively influence goal patterns. Ultimately, we aim to resolve the question: Is there a random distribution of when goals are scored? 2 Data & Methods 2.1 Data Source: Hudl StatsBomb The dataset used in this study consists of event information collected from 3,433 games across 21 different leagues and competitions, including: FIFA World Cup, African Cup of Nations, Premier League, 1. Bundesliga, La Liga, Serie A, Champions League, Major League Soccer (MLS), and National Women’s Soccer League (NWSL). These leagues and competitions span various seasons, and are aggregated and stored by StatsBomb. The basic unit of this data is an event , or actions that happen over the course of a game, including elements such as passes, carries, shots, substitutions, and fouls. This dataset has three partitions: events, lineups, and matches. Detailed information and schema for the StatsBomb data as well as available leagues and competitions can be found at https://github.com/statsbomb/open-data . Events: Each game has a JSON file out of the 3,433 files holding the explicit game data, identified by the match ID. Each game file consists of an array with every event that was recorded in the game as a series of discreet, timestamped events. Each event contains 20 descriptive attributes, including the event timestamp, the type of event (i.e., goal, pass, dribble, etc.), and the players involved. In sum, there are 34 different event types, with “Ball Receipt”, “Shot”, and “Pass” as the most common. The “Shot” event contains additional information on the shot event, such as the start/end location of the shot, the xG (a measure of expectation to score), and the shot’s outcome (e.g. “Goal”, “Blocked”, “Saved”, etc. See Fig. A.1 a). Lineups: Lineups contains information on the lineups and lineup changes for the same 3,433 games. Each file consists of five elements for each player: “ID” (generated by StatsBomb specific to the player), “name”, “nickname”, “jersey number”, and “country/nationality” (See Fig. A.1 b). Matches: Matches contains the specific league or competition (e.g., Premier League) and season (e.g., 2015/2016) for each of the 3,433 matches. It is split into leagues, which then have season inside (See Fig. A.1 c). Each section contains various amounts of JSON files, with each file holding information on matches played during a specific season. The matches folder was used to collect all of the 3,433 match ID’s, and to collect specific match IDs from a league and/or season. 2.2 A null model for goal scoring In order to compare the observed distribution of the timing of (and between) goals, we introduce a null model for estimating the expected number of goals within any given time interval. To construct this goal-scoring null model, we assume that the probability of scoring a goal is consistent across all minutes and matches, defined by the average number of goals per minute, \bar{G} : \bar{G}=\frac{1}{N}\sum_{i=1}^{N}\frac{G_{i}}{T_{i}}, (1) where G_{i} , T_{i} , and N represent the number of goals, the duration in minutes of match i , and the number of matches, respectively. We simulated 6,000,000 games with goals scored at a rate of \bar{G} and game lengths set to the average game length, \bar{T} . In each simulated game, goals were randomly assigned to one of the two teams. 3 Results 3.1 Goal timing On first glance, there does not seem to be an intuitive rationale for why we would expect a goal to be scored at any specific time during a game. This suggests that, by measuring the time that goals are scored across many games, we would expect to see a maximum entropy, uniform distribution across the duration of a game. In Fig. 1 a, we