ECSS Paris 2023: OP-AP03
INTRODUCTION: Professional soccer teams’ season are characterized by contextual factors that could affect players’ perceptual demand. For this reason, this study investigated how weekly internal training load can be modulated by the previous and upcoming competitive context. Specifically, we analysed the effects of the following opponents’ ranking and match location on training impulse (TRIMP) and rating of perceived exertion (RPE), also separating starters and non–starters subgroups. METHODS: Weekly training load data were collected along an entire competitive season in a Serie A soccer team recently relegated to a lower division. Internal load during the week preceding the match was quantified using session–RPE (sRPE) and Banister’s TRIMP, while external training load (ETL) was monitored according with the distance per minute indicator. Opponent current ranking (LOW 14th–20th place, MID 8th–13th place, TOP 1st–7th place) and match location (home vs away) referred to the subsequent official match. Players were classified in starters (who performed at least 30 minutes in both previous and subsequent matches relative to the analysed training week) and non–starters. Linear mixed–effects models were applied to assess: i) differences between starters and non–starers in terms of sRPE and TRIMP, and their relationship; ii) differences in terms of opponent ranking and match location in starters and non–starters for sRPE and TRIMP; iii) the relationship between sRPE and TRIMP considering ranking and match location in starters and non–starters. RESULTS: Starters exhibited lower sRPE (ES=0.2) and TRIMP (ES=0.4) compared with non–starters. Opponent ranking and match location significantly influenced sRPE even after accounting for TRIMP both in two subgroups, indicating a context–dependent dissociation between physiological and perceptual load. Despite weekly TRIMP was significantly higher before matches against TOP opponents in both groups (starters ES=0.2; non–starters ESrange=0.2–0.5), the highest sRPE emerged before matches against LOW (starters ES=0.5; non–starters ES=0.2), TOP opponents (starters ES=0.4; non–starters ES=0.4) and away matches (starters ES=0.3; non–starters ES=0.2). Moreover, the within–player relationship between sRPE and TRIMP was significantly steeper in starters (ß=2.86) than in non–starters (ß=2.63). This relationship was further modulated by opponent ranking, indicating differences in load perception across competitive contexts. CONCLUSION: Weekly internal training load resulted strongly affected by the competitive context. A higher perceptual response against LOW opponents (similar ranking of the team studied) in starters shows a context–sensitive modulation of the physiological–perceptual load relationship. These findings highlight the importance of integrating contextual variables into training load monitoring and support the complementary use of physiological and perceptual parameters to capture distinct components of internal load in elite soccer.
Read CV Damiano Li VolsiECSS Paris 2023: OP-AP03
INTRODUCTION: Field-based running in soccer displays repeated accelerations and decelerations. di Prampero et al. (2005) developed an equivalent slope method to quantify the added energy cost of accelerated and decelerated running (1). Since accelerations can cost up to 5-fold more than steady-state running, this approach better captures metabolic power variations characteristics of team sports. However, total metabolic power does not reveal how energy is supplied across alactic, glycolytic and aerobic pathways. This quantification is relevant to practitioners because athletes with different metabolic profiles deplete energy stores at different rates despite identical external loads, affecting fatigue onset and recovery needs (2). Mader proposed a dynamic model allowing the partition of energy supply during exercise with power variations, based on individual capacities (3). This study explores the integration of GPS- derived metabolic power within Mader’s framework and its practical insights. METHODS: All numerical values are expressed as mean ± standard deviation. Twenty-one male soccer players, with measured maximal oxygen consumption rates (VO2max) of 60.7 ± 6.8 mL/kg/min, completed self-paced 50-m shuttle runs to exhaustion (duration: 19.7 ± 2.1 min; peak speed: 21.1 ± 1.5 km/h), wearing GPS units (GPEXE, 20 Hz) and portable metabolic analyzers (Cosmed K5). Metabolic power was computed from GPS speed and acceleration (1). A five-state ordinary differential equation system modelled muscle VO2, muscle and blood lactate concentrations, and mouth-level VO2 (introducing a time delay relative to muscle VO2) throughout the test. For each athlete, maximal rate of glycolysis (VLamax: 0.54 ± 0.20 mmol/kg/s) and steady-state energy cost of running (4.57 ± 0.36 J/kg/m) were fitted to match experimental end-test blood lactate concentration ([La]) and measured VO2. RESULTS: VO2 agreement showed a root mean square error of 5.3 ± 1.5 mL/kg/min. The differences between simulated and measured total oxygen consumption were 0.4 ± 4.6%. Modelled [La] error was -0.4 ± 1.6 mmol/L. Athletes achieving similar [La] showed VLamax differing up to 4-fold depending on VO2max and running economy. Players with lower VO2max required higher VLamax to reach equivalent [La]. Simulations also indicate that athletes with a higher VLamax deplete their carbohydrate stores more quickly. CONCLUSION: Integrating GPS-derived metabolic power with Mader’s dynamic bioenergetic model reveals insights that initial testing fails to capture. The energy partition helps identify how athletes’ VO2max and VLamax interactions can affect energy use and availability. This information could orientate individualized nutrition and recovery strategies. Future efforts will focus on enhancing parameter-fitting methods and conducting tests across a variety of scenarios, including simulated games. 1) di Prampero et al. J Exp Biol, 2005; 2) Wackerhage et al. Sports Med, 2025; 3) Mader Eur J Appl Physiol, 2003
Read CV Jeremy BriandECSS Paris 2023: OP-AP03
INTRODUCTION: Forecasting how players reposition right after a pass is central to understanding team tactics, judging decisions, and shaping realistic training in soccer[1,2]. Existing approaches often focus on pass difficulty or value (e.g., expected pass) without providing coordinated forecasts of full team formations or individual movements[3,4]. In this work, we propose a transformer model that predicts post-pass locations at three granularities -single player, possession team (11), and both teams (22)- using FIFA World Cup data. The aim is to provide a scalable tool for reconstructing immediate post-pass formations to support performance analysis, scouting, and training design. METHODS: Event and tracking data from matches of the 2022 FIFA World Cup were processed into three-event pass sequences. For single-player models, high-participation French players (>50%) were selected. Sequences encoded 66 features, including player positions (ordered arbitrarily or by formation role), event descriptors, ball coordinates, and pitch-aware spatial indicators. The proposed Pitch-Aware Temporal Transformer model integrates short- and medium-term branches with multi-head attention and LSTMs to predict single-player or team-level coordinates. Baselines trained on 37 non-French matches (80/10/10 split) were compared to models fine-tuned on France's 7 matches and France-only baselines. Performance was quantified using MAE, RMSE, and R2. RESULTS: The transformer consistently improved forecasts, with benefits from role-ordering and fine-tuning. For unordered inputs, fine-tuning reduced both-teams MAE from 8.1 m to 5.8 m (28% reduction) and possession-team MAE from 6.3 m to 4.7 m (25% reduction). Fine-tuning outperformed direct France-only training by 14-15%, indicating that adapting a multi-team model yields better tactical forecasts than training from scratch. With ordered inputs, fine-tuning improved both-teams MAE to 5.1 m (R2 0.85) and possession-team MAE to 4.8 m (R2 0.87). Ordered variants outperformed unordered by approximately 10-15%. The single-player model achieved the highest precision (MAE 3.4 m, R2 0.9, RMSE 4.8 m), improving 29-37% over team-level models. CONCLUSION: The proposed transformer model accurately reconstructs post-pass formations, particularly when player roles are ordered and models are fine-tuned. Notably, our single-player model (MAE 3.4 m, RMSE 4.8 m) outperforms previously reported related tasks such as Agent Imputer (MAE 4.5 m and RMSE 6.9 m) [5]. Robust gains at the team level highlight the framework's ability to capture coordinated tactical reorganization. Unlike approaches that only estimate pass value, this work delivers a variable-granularity tool to simulate likely post-pass configurations. For sports scientists and coaches, this offers a scalable way to link event data to realistic on-pitch reorganization, supporting informed decision-making in scouting and session planning.References: [1]Teixeira2025 /[2]Ashford2021 /[3]Fernandez2021 /[4]StatsBomb2023 /[5]Everett2023
Read CV Mikhail AbdelmalakECSS Paris 2023: OP-AP03