The present thesis is based on 14 original articles published in international journals (I-XIV, see page 8) and a summarizing review. The thesis deals with the physiological demands of soccer, with a particular focus on the physiological response to repeated intense exercise. In chapter I the specific issues are presented and in chapter II the physiological demands in soccer are discussed based on the results of the studies performed. Chapter III contains a short survey of the experiments performed to study specifically muscle metabolism and muscle fatigue with repeated intense muscle contractions. With reference to the topics covered in chapters II and III, fatigue during a soccer match is discussed in chapter IV, and chapter V deals with applications for physical training in soccer. Measurements have been performed during soccer matches and training, as well as in experiments simulating the activities of a soccer match. The information obtained has been compared to results from studies of the physical capacity of top-class soccer players and from laboratory experiments aimed at investigating metabolism and fatigue in intermittent exercise. Studies with whole-body and single muscle group exercises have been performed, the latter mainly with the application of a knee-extension model. In the studies on isolated muscle groups, biopsies taken from exercising muscles as well as arterial and femoral venous blood samples have allowed for detailed analysis of muscle ionic transportation and metabolism. In addition, the magnetic resonance technique has been used for the continuous determination of changes in muscle metabolites and pH during intermittent exercise. Analysis of activities during soccer matches showed that a top-class soccer player covers an average distance of approximately 11 km during a match. The distance differs highly between players and is partly related to the position in a team. Midfield players run more at low speed than defenders and forwards, whereas no difference appears to exist between groups when comparing the distance covered at high speed. The distance covered at high speed is the same in the beginning as in the end of a match. The total distance covered by a player during a soccer match is only to a limited extent a measure of the physiological demands on the player during the match. In addition to running, a player is engaged in many other energy demanding activities, i.e. tackling, jumping, accelerating and turning. A more precise evaluation of the total energy demand during a soccer match may be achieved by performing physiological measurements in connection with soccer matches (I).(ABSTRACT TRUNCATED AT 400 WORDS)
In order to study the movement pattern of soccer players, 14 top-level players were filmed during several competitive matches. In addition, the relationship between the observed activity during match play and blood lactate values was examined. The mean distance covered during competitive matches was 10.80 km, and the average individual difference between matches was 0.92 km, with no difference in regard to high intensity activities. Midfielders covered a 10% longer (p less than 0.05) distance (11.4 km) than defenders and forwards, with no difference concerning high intensity running. There was a significant correlation (r = 0.61, p less than 0.05) between the amount of high intensity running during the match and lactate concentration in the blood. The results suggest that high intensity running can be used for making comparisons in soccer and that the interpretation of blood lactate in soccer is limited to giving an indication of the type of activity that has been carried out a few minutes before sampling.
The purpose of this study was to evaluate the physical demands of English Football Association (FA) Premier League soccer of three different positional classifications (defender, midfielder and striker). Computerised time-motion video-analysis using the Bloomfield Movement Classification was undertaken on the purposeful movement (PM) performed by 55 players. Recognition of PM had a good inter-tester reliability strength of agreement (κ= 0.7277). Players spent 40.6 ± 10.0% of the match performing PM. Position had a significant influence on %PM time spent sprinting, running, shuffling, skipping and standing still (p < 0.05). However, position had no significant influence on the %PM time spent performing movement at low, medium, high or very high intensities (p > 0.05). Players spent 48.7 ± 9.2% of PM time moving in a directly forward direction, 20.6 ± 6.8% not moving in any direction and the remainder of PM time moving backward, lateral, diagonal and arced directions. The players performed the equivalent of 726 ± 203 turns during the match; 609 ± 193 of these being of 0° to 90° to the left or right. Players were involved in the equivalent of 111 ± 77 on the ball movement activities per match with no significant differences between the positions for total involvement in on the ball activity (p > 0.05). This study has provided an indication of the different physical demands of different playing positions in FA Premier League match-play through assessment of movements performed by players. Key pointsPlayers spent ~40% of the match performing Pur-poseful Movement (PM).Position had a significant influence on %PM time spent performing each motion class except walking and jogging. Players performed >700 turns in PM, most of these being of 0°-90°.Strikers performed most high to very high intensity activity and most contact situations.Defenders also spent a significantly greater %PM time moving backwards than the other two posi-tions.Different positions could benefit from more specific conditioning programs.
Recent advances in soccer understanding have demonstrated rapid progress, yet existing research predominantly focuses on isolated or narrow tasks. To bridge this gap, we propose a comprehensive framework for holistic soccer understanding. Concretely, we make the following contributions in this paper: (i) we construct SoccerWiki, the first large-scale multimodal soccer knowledge base, integrating rich domain knowledge about players, teams, referees, and venues to enable knowledge-driven reasoning; (ii) we present SoccerBench, the largest and most comprehensive soccer-specific benchmark, featuring around 10K multimodal (text, image, video) multi-choice QA pairs across 13 distinct tasks; (iii) we introduce SoccerAgent, a novel multi-agent system that decomposes complex soccer questions via collaborative reasoning, leveraging domain expertise from SoccerWiki and achieving robust performance; (iv) extensive evaluations and comparisons with representative MLLMs on SoccerBench highlight the superiority of our agentic system.
Soccer is a globally popular sporting event, typically characterized by long matches and distinctive highlight moments. Recent advances in Multimodal Large Language Models (MLLMs) offer promising capabilities in temporal grounding and video understanding, soccer commentary generation often requires precise temporal localization and semantically rich descriptions over long-form video. However, existing soccer MLLMs often rely on the temporal a priori for caption generation, so they cannot process the soccer video end-to-end. While some traditional approaches follow a two-step paradigm that is complex and fails to capture the global context to achieve suboptimal performance. To solve the above issues, we present TimeSoccer, the first end-to-end soccer MLLM for Single-anchor Dense Video Captioning (SDVC) in full-match soccer videos. TimeSoccer jointly predicts timestamps and generates captions in a single pass, enabling global context modeling across 45-minute matches. To support long video understanding of soccer matches, we introduce MoFA-Select, a training-free, motion-aware frame compression module that adaptively selects representative frames via a coarse-to-fine strategy, and in
The integration of artificial intelligence in sports analytics has transformed soccer video understanding, enabling real-time, automated insights into complex game dynamics. Traditional approaches rely on isolated data streams, limiting their effectiveness in capturing the full context of a match. To address this, we introduce SoccerChat, a multimodal conversational AI framework that integrates visual and textual data for enhanced soccer video comprehension. Leveraging the extensive SoccerNet dataset, enriched with jersey color annotations and automatic speech recognition (ASR) transcripts, SoccerChat is fine-tuned on a structured video instruction dataset to facilitate accurate game understanding, event classification, and referee decision making. We benchmark SoccerChat on action classification and referee decision-making tasks, demonstrating its performance in general soccer event comprehension while maintaining competitive accuracy in referee decision making. Our findings highlight the importance of multimodal integration in advancing soccer analytics, paving the way for more interactive and explainable AI-driven sports analysis. https://github.com/simula/SoccerChat
Vision-language models (VLMs) have recently shown strong potential in soccer video understanding. However, given the high complexity of soccer videos due to large viewpoint variations, rapid shot transitions, and cluttered scenes, it remains unclear on whether VLMs rely on meaningful visual evidence or exploit spurious correlations and shortcut learning. Existing evaluation protocols focus primarily on classification accuracy and do not assess visual grounding. To address this limitation, we introduce SoccerLens, a benchmark for grounded soccer video understanding. The benchmark contains annotated video segments spanning $13$ common soccer events, with structured visual cues organized into three levels of semantic relevance. We further extend the attribution method of Chefer [arXiv:2103.15679] to jointly model spatial and temporal attention, and introduce evaluation metrics that measure whether model attention aligns with annotated cues or drifts toward spurious regions. Our evaluation of state-of-the-art soccer VLMs shows that, despite strong classification accuracy, current models fail to exceed $50\%$ grounding performance even under the loosest cue definitions and consistently
Soccer understanding has recently garnered growing research interest due to its domain-specific complexity and unique challenges. Unlike prior works that typically rely on isolated, task-specific expert models, this work aims to propose a unified model to handle diverse soccer visual understanding tasks, ranging from fine-grained perception (e.g., athlete detection and identification) to high-level semantic reasoning (e.g., event classification). Concretely, our contributions are threefold: (i) we present SoccerMaster, the first soccer-specific vision foundation model that unifies diverse tasks within a single framework via supervised multi-task pretraining; (ii) we develop an automated data curation pipeline, SoccerFactory, to generate scalable spatial annotations, and integrate multiple existing soccer video datasets as a comprehensive pretraining data resource for multi-task pretraining; and (iii) we conduct extensive evaluations demonstrating that SoccerMaster consistently outperforms task-specific expert models across diverse downstream tasks, highlighting its breadth and superiority. The data, code, and model will be publicly available.
Soccer is a globally renowned sport with significant applications in video games and VR/AR. However, generating realistic soccer motions remains challenging due to the intricate interactions between the human player and the ball. In this paper, we introduce SMGDiff, a novel two-stage framework for generating real-time and user-controllable soccer motions. Our key idea is to integrate real-time character control with a powerful diffusion-based generative model, ensuring high-quality and diverse output motion. In the first stage, we instantly transform coarse user controls into diverse global trajectories of the character. In the second stage, we employ a transformer-based autoregressive diffusion model to generate soccer motions based on trajectory conditioning. We further incorporate a contact guidance module during inference to optimize the contact details for realistic ball-foot interactions. Moreover, we contribute a large-scale soccer motion dataset consisting of over 1.08 million frames of diverse soccer motions. Extensive experiments demonstrate that our SMGDiff significantly outperforms existing methods in terms of motion quality and condition alignment.
Vision Language Models (VLMs) have demonstrated strong performance in multi-modal tasks by effectively aligning visual and textual representations. However, most video understanding VLM research has been domain-agnostic, leaving the understanding of their transfer learning capability to specialized domains under-explored. In this work, we address this by exploring the adaptability of open-source VLMs to specific domains, and focusing on soccer as an initial case study. Our approach uses large-scale soccer datasets and LLM to create instruction-following data, and use them to iteratively fine-tune the general-domain VLM in a curriculum learning fashion (first teaching the model key soccer concepts to then question answering tasks). The final adapted model, trained using a curated dataset of 20k video clips, exhibits significant improvement in soccer-specific tasks compared to the base model, with a 37.5% relative improvement for the visual question-answering task and an accuracy improvement from 11.8% to 63.5% for the downstream soccer action classification task.
This paper introduces an approach to predicting the next event in a soccer match, a challenge bearing remarkable similarities to the problem faced by Large Language Models (LLMs). Unlike other methods that severely limit event dynamics in soccer, often abstracting from many variables or relying on a mix of sequential models, our research proposes a novel technique inspired by the methodologies used in LLMs. These models predict a complete chain of variables that compose an event, significantly simplifying the construction of Large Event Models (LEMs) for soccer. Utilizing deep learning on the publicly available WyScout dataset, the proposed approach notably surpasses the performance of previous LEM proposals in critical areas, such as the prediction accuracy of the next event type. This paper highlights the utility of LEMs in various applications, including match prediction and analytics. Moreover, we show that LEMs provide a simulation backbone for users to build many analytics pipelines, an approach opposite to the current specialized single-purpose models. LEMs represent a pivotal advancement in soccer analytics, establishing a foundational framework for multifaceted analytics p
As a globally celebrated sport, soccer has attracted widespread interest from fans all over the world. This paper aims to develop a comprehensive multi-modal framework for soccer video understanding. Specifically, we make the following contributions in this paper: (i) we introduce SoccerReplay-1988, the largest multi-modal soccer dataset to date, featuring videos and detailed annotations from 1,988 complete matches, with an automated annotation pipeline; (ii) we present an advanced soccer-specific visual encoder, MatchVision, which leverages spatiotemporal information across soccer videos and excels in various downstream tasks; (iii) we conduct extensive experiments and ablation studies on event classification, commentary generation, and multi-view foul recognition. MatchVision demonstrates state-of-the-art performance on all of them, substantially outperforming existing models, which highlights the superiority of our proposed data and model. We believe that this work will offer a standard paradigm for sports understanding research.
We propose a foundation model for soccer, which is able to predict subsequent actions in a soccer match from a given input sequence of actions. As a proof of concept, we train a transformer architecture on three seasons of data from a professional soccer league. We quantitatively and qualitatively compare the performance of this transformer architecture to two baseline models: a Markov model and a multi-layer perceptron. Additionally, we discuss potential applications of our model. We provide an open-source implementation of our methods at https://github.com/danielhocevar/Foundation-Model-for-Soccer.
In this paper, we present a novel sequential team selection model in soccer. Specifically, we model the stochastic process of player injury and unavailability using player-specific information learned from real-world soccer data. Monte-Carlo Tree Search is used to select teams for games that optimise long-term team performance across a soccer season by reasoning over player injury probability. We validate our approach compared to benchmark solutions for the 2018/19 English Premier League season. Our model achieves similar season expected points to the benchmark whilst reducing first-team injuries by ~13% and the money inefficiently spent on injured players by ~11% - demonstrating the potential to reduce costs and improve player welfare in real-world soccer teams.
RoboCup Soccer Simulation 2D (SS2D) research is hampered by the complexity of existing Cpp-based codes like Helios, Cyrus, and Gliders, which also suffer from limited integration with modern machine learning frameworks. This development paper introduces a transformative solution a gRPC-based, language-agnostic framework that seamlessly integrates with the high-performance Helios base code. This approach not only facilitates the use of diverse programming languages including CSharp, JavaScript, and Python but also maintains the computational efficiency critical for real time decision making in SS2D. By breaking down language barriers, our framework significantly enhances collaborative potential and flexibility, empowering researchers to innovate without the overhead of mastering or developing extensive base codes. We invite the global research community to leverage and contribute to the Cross Language Soccer (CLS) framework, which is openly available under the MIT License, to drive forward the capabilities of multi-agent systems in soccer simulations.
Camera calibration and localization, sometimes simply named camera calibration, enables many applications in the context of soccer broadcasting, for instance regarding the interpretation and analysis of the game, or the insertion of augmented reality graphics for storytelling or refereeing purposes. To contribute to such applications, the research community has typically focused on single-view calibration methods, leveraging the near-omnipresence of soccer field markings in wide-angle broadcast views, but leaving all temporal aspects, if considered at all, to general-purpose tracking or filtering techniques. Only a few contributions have been made to leverage any domain-specific knowledge for this tracking task, and, as a result, there lacks a truly performant and off-the-shelf camera tracking system tailored for soccer broadcasting, specifically for elevated tripod-mounted cameras around the stadium. In this work, we present such a system capable of addressing the task of soccer broadcast camera tracking efficiently, robustly, and accurately, outperforming by far the most precise methods of the state-of-the-art. By combining the available open-source soccer field detectors with ca
Achieving coordinated teamwork among legged robots requires both fine-grained locomotion control and long-horizon strategic decision-making. Robot soccer offers a compelling testbed for this challenge, combining dynamic, competitive, and multi-agent interactions. In this work, we present a hierarchical multi-agent reinforcement learning (MARL) framework that enables fully autonomous and decentralized quadruped robot soccer. First, a set of highly dynamic low-level skills is trained for legged locomotion and ball manipulation, such as walking, dribbling, and kicking. On top of these, a high-level strategic planning policy is trained with Multi-Agent Proximal Policy Optimization (MAPPO) via Fictitious Self-Play (FSP). This learning framework allows agents to adapt to diverse opponent strategies and gives rise to sophisticated team behaviors, including coordinated passing, interception, and dynamic role allocation. With an extensive ablation study, the proposed learning method shows significant advantages in the cooperative and competitive multi-agent soccer game. We deploy the learned policies to real quadruped robots relying solely on onboard proprioception and decentralized localiz
In soccer, game context can result in skewing offensive statistics in ways that might misrepresent how well a team has played. For instance, in England's 1-2 loss to France in the 2022 FIFA World Cup quarterfinal, England attempted considerably more shots (16 to France's 8) and more corners (5 to 2), potentially suggesting they played better despite the loss. However, these statistics were largely accumulated when France was ahead and more willing to concede offensive initiative to England. To explore how game context influences offensive performance, we analyze minute-by-minute event-sequenced match data from 15 seasons across five major European leagues. Using count-response Generalized Additive Modeling, we consider features such as score and red card differential, home/away status, pre-match win probabilities, and game minute. Moreover, we leverage interaction terms to test several intuitive hypotheses about how these features might cooperate in explaining offensive production. The selected model is then applied to project offensive statistics onto a standardized "common denominator" scenario: a tied home game with even men on both sides. The adjusted numbers - in contrast to r
The rapid evolution of digital sports media necessitates sophisticated information retrieval systems that can efficiently parse extensive multimodal datasets. This paper introduces SoccerRAG, an innovative framework designed to harness the power of Retrieval Augmented Generation (RAG) and Large Language Models (LLMs) to extract soccer-related information through natural language queries. By leveraging a multimodal dataset, SoccerRAG supports dynamic querying and automatic data validation, enhancing user interaction and accessibility to sports archives. Our evaluations indicate that SoccerRAG effectively handles complex queries, offering significant improvements over traditional retrieval systems in terms of accuracy and user engagement. The results underscore the potential of using RAG and LLMs in sports analytics, paving the way for future advancements in the accessibility and real-time processing of sports data.
The rapid evolution of digital sports media necessitates sophisticated information retrieval systems that can efficiently parse extensive multimodal datasets. This paper demonstrates SoccerRAG, an innovative framework designed to harness the power of Retrieval Augmented Generation (RAG) and Large Language Models (LLMs) to extract soccer-related information through natural language queries. By leveraging a multimodal dataset, SoccerRAG supports dynamic querying and automatic data validation, enhancing user interaction and accessibility to sports archives. We present a novel interactive user interface (UI) based on the Chainlit framework which wraps around the core functionality, and enable users to interact with the SoccerRAG framework in a chatbot-like visual manner.