The quantification of external training load and its association with match demands is a critical focus for sports with high-intensity demands such as soccer. Understanding the external load imposed on athletes during training and matches is essential for optimizing performance and ensuring recovery (Halson, 2014). External load refers to the physical demands imposed on players, typically measured using technologies such as global positioning systems (GPS) with integrated inertial measurement units (IMUs) and magnetometers. These devices, manufactured by companies, such as Catapult Sports, GPSports, StatSports, and Polar, have become standard tools for quantifying movement patterns, kinematic variables, and mechanical loading in team sports (Rago et al., 2020). GPS technology provides spatial and velocity-based metrics (e.g., total distance [TD], high-speed running [HSR]), while integrated tri-axial accelerometers capture acceleration (ACC)-based measures (e.g., player load [PL], ACCs/decelerations [DECs]), and IMUs quantify changes in body orientation and direction (Malone et al., 2017). Complementary camera-based tracking systems (e.g., TRACAB, ChyronHego) offer alternative approaches for capturing positional and tactical data, though GPS remains the most widely adopted method for individual PL monitoring in soccer (Rago et al., 2020). Monitoring these measures allows coaches and practitioners to ensure that training loads are appropriately individualized and have proximity to the match demands, thus improving performance (Foster et al., 2017).
In soccer, match demands can be highly variable depending on factors such as the player’s position, the intensity of the match, and the tactical strategies employed (Modric et al., 2023). Soccer players are required to perform a wide range of activities, from HSR and sprinting to rapid changes in direction and ACC/DEC cycles (Gualtieri et al., 2023). As such, understanding the external load demands during a match is crucial for training program that adequately prepares athletes for competition. The interest in understanding training and matching external load demands in female soccer has grown considerably (Pérez Armendáriz et al., 2024). Female soccer players experience physical demands that, while structurally similar to male soccer, can vary in intensity, frequency, and physiological response (Winther et al., 2024; Randell et al., 2021). Research has shown that players in women’s soccer cover substantial TDs, engage in frequent ACCs and DECs, and perform HSR and sprinting actions during competitive matches (Winther et al., 2024; Pérez Armendáriz et al., 2024; Vescovi et al., 2021). However, training methodologies often remain generalized and not always programmed to the specific demands experienced by female players during competition (Romero-Moraleda et al., 2021). Therefore, it is critical to analyze and align the training load of different training session types with actual match demands to ensure optimal performance, injury prevention, and athlete development in female soccer (Vescovi et al., 2021).
While match analysis provides valuable information regarding the external load demands of the sport, there is a need for systematic methods to measure the external load during training that closely replicates the intensity and volume of match play. Given that, small-sided games (SSGs), conditioning (C), and tactical (T) training sessions play a significant role in this context. SSGs effectively mimic match-specific scenarios while allowing for the manipulation of training intensity, volume, and tactical complexity (Halouani et al., 2014). Previous studies have shown that SSGs are an effective method for replicating the external load demands of match play, particularly in terms of high-intensity running, ACCs, and DECs (Clemente et al., 2022; Mandorino et al., 2024). However, while SSGs are often considered to replicate match intensity, there remains debate on their ability to fully capture the diversity of match demands, particularly for certain external load variables such as HSR, and sprinting (Clemente, 2020).
Conditioning sessions that focus on improving specific physical capacities, such as ACC, DEC, and sprinting alone, are also widely used in soccer training (García-Ramos et al., 2018). These sessions are typically designed to focus on high-intensity running and repeated sprint efforts (Clemente et al., 2021), and are structured to improve aerobic and anaerobic capacity, making them particularly valuable for addressing the high-intensity demands of match play (Clemente et al., 2021). However, conditioning sessions tend to be more structured and less dynamic than SSGs, and their ability to replicate match-specific activities such as tactical decisions or team interactions is limited (Arslan et al., 2020). Despite this, conditioning sessions have shown strong correlations with external load measures, such as ACCs and DECs, which are key components of match performance (Harper et al., 2019). For such reasons, exploring the relationships between the different types of training session demands and match demands allows for improving training programs that target the most relevant external load demands for greater match readiness.
The relationship between training load and match load is an essential area of investigation, as it can provide important information regarding the training dose for players (Fitzpatrick et al., 2018). Previous research has shown that the external load during training can have a significant impact on match performance, with higher training loads often leading to improved performance outcomes in matches (Silva et al., 2021, 2023). However, the training-to-match ratio, which simply divides the training load by the match load, can vary significantly depending on the type of training conducted (Silva et al., 2023). SSGs provide a higher training-to-match ratio for external load measures, while conditioning sessions may be more effective for replicating ACC and DEC demands (Savolainen et al., 2023). Understanding these relationships is crucial for coaches, as it allows them to manipulate training loads to match the specific requirements of upcoming matches. As soccer matches involve complex, dynamic, and tactical interactions, it is essential to explore whether different training types can adequately replicate all aspects of the external load seen during a match. The aim of this study is to determine which training session types (SSGs, tactical, conditioning, or combined formats) best replicate the specific external load measures observed in match play, providing coaches with evidence-based guidance for training session design that optimally prepares elite female soccer players for competition demands.
An observational, analytical cohort design was used to collect the data. A female professional soccer national team was observed for 6 weeks (from 7 August 2024 to 18 September 2024). Each player was observed during all sessions and official matches using a GPS unit. During each training week, the coaching staff implemented their training process consisting of four different types of sessions: (i) conditioning only; (ii) SSGs only; (iii) tactical only; and (iv) all combined (conditioning, SSGs, and tactical, C + SSGs + T). Accordingly, this study investigated the associations between the external load measures recorded during each training session type and the external load recorded during subsequent matches. A standardized warm-up was included in all sessions.
All athletes were monitored during the observational period using a 15 Hz GPS unit with a 100 Hz tri-axial accelerometer, a 50 Hz magnetometer, and a 16 G tri-axial impact tracker from a GPSports System (Canberra, Australia). The GPSports System was previously considered valid and reliable for measuring distances in tennis players [18]. Each player used his attributed GPS unit during both training and matches. For each session recorded, these were the measures extracted: (i) TD; (ii) PL (forces quantified by a 100 Hz tri-axial accelerometer); (iii) ACCs (>4.0 m/s2); (iv) DECs (>4.0 m/s2); (v) HSR (>19.8 km/h); (vi) MaxSpeed; (vii) RDs, quantify changes in running direction and represent the number of directional changes per minute, reflecting the multidirectional nature of movements during training and matches; and (viii) FS represent the total number of ground contacts per minute, measured via the tri-axial accelerometer, and serve as an indicator of locomotor frequency and neuromuscular load.
The weekly mean training intensity for each external load measure was considered for the analysis. That is, the mean number of actions/meters covered during a training week by each athlete was relativized per minute. Then, all the analyzed external load measures were relativized per minute to quantify the match intensity.
Twenty-four female soccer players (age: 24.5 ± 2.3 years old; height: 171 ± 3.0 cm; body mass: 63.5 ± 4.0 kg), from the Turkish National Team participated in this study. The inclusion criteria were based on: (i) being present at all sessions and the match of the same week for all weeks and (ii) not being injured in the 2 weeks preceding match participation. Players who were absent for more than 2 weeks were excluded from the sample. The study commenced following approval from the Ethics Committee of the Instituto Politécnico de Viana do Castelo (Reference: CECSVS2024/02/vi). All participants were fully informed about the study’s objectives, procedures, and risks, and provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki.
The conditioning sessions commenced with a standardized warmup lasting approximately 10–15 min, including dynamic stretching, mobility drills, and light aerobic jogging at 60–70% of maximum heart rate (HRmax). High-intensity interval training (HIIT) was integrated, characterized by short bursts of maximal effort followed by brief recovery periods. HIIT involved 4–6 sets of 30-s all-out sprints and shuttle runs at 90–100% HRmax, with 60–90 s of active recovery at 50–60% HRmax, totaling 15–20 min per session. Speed interval training (SIT), categorized as extensive (Ext SIT) or intensive (Int SIT), was further implemented. Ext SIT included 8–10 repetitions of 60 m runs at 70–80% of MaxSpeed with 60 s of rest, lasting 15–20 min, while Int SIT consisted of 6–8 repetitions of 30 m sprints at 90–95% of MaxSpeed with 90 s of rest, lasting 10–15 min.
SSG sessions implemented 1v1, 2v2, 3v3 formats, and larger formats such as 7v7 with neutral players or goalkeepers, alongside 8v8 and 10v10 possession exercises, rondos, and pressing formats such as 8v4 and 12v6 presses. SSGs incorporated possession retention, rapid passing circles, and defensive pressure exercises. Each SSG format was conducted on pitches ranging from 20 × 30 m for 1v1 to 60 × 40 m for 7v7, with sessions lasting 45–60 min. For example, 3v3 games typically involved 4–6 bouts of 3–4 min at high intensity (80–90% HRmax) with 2-min rest intervals, while 8v8 possession exercises included 3–4 bouts of 6–8 min with 3-min rest periods. Rondos (e.g., 5v2) were performed for 10–15 min in 2–3 sets, emphasizing quick passing and movement. Pressing formats, such as 8v4, involved 3–5 bouts of 2–3 min at 85–95% HRmax to simulate high-pressure match scenarios. All sessions were planned and implemented by the coaching staff and concluded with a 5–10 min cool-down of light jogging and stretching, resulting in total session durations of 60–90 min.
Tactical sessions implemented set piece practice, 11v11 full-team tactical rehearsal exercises, three-back pressing structures, and tactical exercises with smaller groups scaling to full-team tactical play. Activities included free kicks, corners, throw-ins, formation rehearsals, pressing strategies, offensive transition runs, and pattern-of-play exercises, scheduled frequently before matches. Set piece practice, such as free kicks and corners, typically lasted 15–20 min, with 10–12 repetitions per scenario at moderate intensity (60–70% HRmax). The 11v11 tactical rehearsals, conducted on a full-size pitch, lasted 20–30 min and focused on specific formations (e.g., 4-3-3 or 3-5-2) and game scenarios, such as build-up play or defensive transitions, at 70–80% HRmax. Three-back pressing structures involved 10–15 min of drills in groups of 6–8 players, scaling to full-team exercises, with 3–4 bouts of 3–5 min at 80–85% HRmax. Pattern-of-play exercises, emphasizing offensive transitions, were conducted for 15–20 min in 2–3 sets of 5–7 min, simulating match-like movements at 75–85% HRmax. All sessions were planned and implemented by the coaching staff and typically lasted 70–90 min, including a 10–15 min standardized warm-up and a 5–10 min cool-down with tactical debriefing.
Combined sessions implemented passing exercises, SSGs, HIIT, Ext SIT, and Int SIT, set piece practice, and 11v11 tactical exercises. Sessions progressed from possession games to pressing scenarios, interval runs, and tactical rehearsals, conducted regularly across the training microcycle. Sessions began with a 10–15 min standardized warm-up (dynamic stretching and light jogging at 60–70% HRmax), followed by 10–15 min of passing exercises (e.g., 4v2 rondos or short-passing circuits) at 65–75% HRmax. SSGs, such as 4v4 or 6v6 on 30 × 40 m pitches, lasted 15–20 min with 3–4 bouts of 4–5 min at 80–90% HRmax and 2-min rest intervals. HIIT involved 10–15 min of 4–6 sets of 30-s maximal efforts (e.g., shuttle runs) at 90–100% HRmax with 60-s recovery. Ext SIT and Int SIT followed the same structure as conditioning sessions, lasting 10–15 min each. Set piece practice (e.g., corners or free kicks) and 11v11 tactical rehearsals each occupied 15–20 min, with intensities of 60–80% HRmax, focusing on match-specific scenarios such as counter-attacks or high pressing. Sessions concluded with a 5–10 min cool-down of light jogging and stretching, resulting in total durations of 90–120 min. All session types were planned and implemented by the team coaching staff and could not be manipulated by the researchers.
Tests of normal distribution and homogeneity (Kolmogorov–Smirnov and Levene’s, respectively) were conducted on all data before analysis. The Spearman’s correlation coefficient was used to examine the relationships between the different training types and the subsequent match external load measures. To interpret the magnitude of these correlations, the following criteria were adopted (Hopkins et al., 2009): r ≤ 0.1, trivial; 0.1 < r ≤ 0.3, small; 0.3 < r ≤ 0.5, moderate; 0.5 < r ≤ 0.7, large; 0.7 < r ≤ 0.9, very large; and r > 0.9, almost perfect. For variables with statistically significant correlations, regression analysis was conducted to explore predictive relationships. All regression analyses were univariate, with each model including a single predictor variable to ensure adequate statistical power given the sample size of 24 subjects. All analyses were performed using JASP software (Version 0.19.3), with a significance threshold of p < 0.05.
The weekly training, mean training intensity, and match demands are listed in Table 1.
Weekly training intensity and match intensity external load
| TD (m/min) | PL (AU/min) | MaxSpeed (km/h/min) | HSR (m/min) | ACC (m/min) | DEC (m/min) | RD (n/min) | FS (n/min) | |
|---|---|---|---|---|---|---|---|---|
| Conditioning | 76.0 ± 8.6 | 8.7 ± 1.7 | 0.3 ± 0.0 | 1.7 ± 0.8 | 0.1 ± 0.0 | −0.1 ± 0.0 | 0.3 ± 0.1 | 37.9 ± 9.1 |
| SSGs | 69.1 ± 8.4 | 7.9 ± 1.1 | 0.3 ± 0.0 | 1.6 ± 1.5 | 0.1 ± 0.0 | −0.1 ± 0.0 | 0.3 ± 0.1 | 21.7 ± 13.1 |
| Tactical | 70.1 ± 11.9 | 7.4 ± 1.4 | 0.4 ± 0.1 | 0.7 ± 0.4 | 0.1 ± 0.0 | −0.1 ± 0.0 | 0.2 ± 0.1 | 22.5 ± 17.3 |
| C+SSGs+T | 78.0 ± 6.8 | 8.0 ± 1.1 | 0.3 ± 0.0 | 4.1 ± 1.0 | 0.1 ± 0.0 | −0.1 ± 0.0 | 0.1 ± 0.0 | 22.3 ± 1.2 |
| Match | 105.1 ± 23.2 | 9.0 ± 2.6 | 0.7 ± 0.7 | 3.5 ± 3.5 | 0.1 ± 0.1 | −0.1 ± 0.1 | 0.2 ± 0.1 | 35.5 ± 25.6 |
FS: footstrikes; TD: total distance; MaxSpeed: maximum speed; HSR: high-speed running; PL: playerload; RD: running deviation; ACC: acceleration; DEC: deceleration; SSGs: small-sided games; C + SSGs + T: conditioning + small-sided games + tactical.
The correlations between conditioning training sessions’ external load and match external load are presented below in Table 2. Significant moderate positive correlations were observed between conditioning PL/min and match HSR/min (r = 0.314, p = 0.036), as well as between conditioning PL/min and match ACC/min (r = 0.300, p = 0.046). Conversely, a significant negative correlation was found between conditioning PL/min and match DEC/min (r = −0.303, p = 0.043). Additionally, significant positive correlations were identified between conditioning HSR/min and match PL/min (r = 0.331, p = 0.026), and between conditioning FS /min and multiple match variables, including TD/min (r = 0.363, p = 0.015), HSR/min (r = 0.420, p = 0.004), ACC/min (r = 0.297, p = 0.048), RD/min (r = 0.297, p = 0.048), and FS /min (r = 0.352, p = 0.018). A significant negative correlation was also found between conditioning RD/min and match HSR/min (r = −0.379, p = 0.011).
Correlations between conditioning sessions and match external load
| Conditioning/match | TD/min | PL/min | MaxSpeed/min | HSR/min | ACC/min | DEC/min | RD/min | FS/min |
|---|---|---|---|---|---|---|---|---|
| TD/min | r = 0.191 | r = 0.237 | r = −0.039 | r = −0.115 | r = 0.031 | r = −0.036 | r = 0.130 | r = −0.101 |
| p = 0.209 | p = 0.117 | p = 0.797 | p = 0.449 | p = 0.842 | p = 0.813 | p = 0.395 | p = 0.508 | |
| PL/min | r = 0.283 | r = 0.287 | r = 0.253 | r = 0.314* | r = 0.300* | r = −0.303* | r = 0.238 | r = 0.144 |
| p = 0.060 | p = 0.056 | p = 0.094 | p = 0.036 | p = 0.046 | p = 0.043 | p = 0.116 | p = 0.345 | |
| MaxSpeed/min | r = 0.066 | r = 0.259 | r = 0.106 | r = 0.181 | r = 0.122 | r = −0.156 | r = 0.157 | r = −0.072 |
| p = 0.665 | p = 0.085 | p = 0.485 | p = 0.234 | p = 0.422 | p = 0.307 | p = 0.301 | p = 0.638 | |
| HSR/min | r = −0.103 | r = 0.331* | r = −0.081 | r = −0.178 | r = −0.093 | r = 0.131 | r = −0.284 | r = 0.107 |
| p = 0.498 | p = 0.026 | p = 0.597 | p = 0.241 | p = 0.544 | p = 0.390 | p = 0.059 | p = 0.483 | |
| ACC/min | r = 0.057 | r = 0.095 | r = 0.142 | r = 0.086 | r = 0.153 | r = −0.167 | r = 0.240 | r = −0.198 |
| p = 0.710 | p = 0.537 | p = 0.351 | p = 0.575 | p = 0.313 | p = 0.272 | p = 0.113 | p = 0.192 | |
| DEC/min | r = −0−009 | r = −0.175 | r = −0.077 | r = −0.090 | r = −0.106 | r = 0.122 | r = −0.179 | r = 0.128 |
| p = 0.952 | p = 0.252 | p = 0.616 | p = 0.555 | p = 0.488 | p = 0.426 | p = 0.239 | p = 0.400 | |
| RD/min | r = −0.237 | r = 0.023 | r = −0.170 | r = −0.379* | r = −0.142 | r = 0.108 | r = −0.130 | r = −0.292 |
| p = 0.117 | p = 0.883 | p = 0.263 | p = 0.011 | p = 0.352 | p = 0.479 | p = 0.394 | p = 0.052 | |
| FS/min | r = 0.363* | r = −0.033 | r = 0.274 | r = 0.420* | r = 0.297* | r = −0.288 | r = 0.297* | r = 0.0352* |
| p = 0.015 | p = 0.831 | p = 0.069 | p = 0.004 | p = 0.048 | p = 0.055 | p = 0.048 | p = 0.018 |
TD: total distance; PL: Playerload; MaxSpeed: maximal speed; HSR: high-speed running; ACC: acceleration; DEC: deceleration; RD: running deviation. SSGs: small-sided games; *: denotes significance at p ≤ 0.05.
The correlations between SSG training sessions’ external load and match external load are presented in Table 3. Significant moderate positive correlations were observed between SSG PL/min and match HSR/min (r = 0.415, p < 0.001), and between SSG DECs/min and match FS /min (r = 0.300, p = 0.017). Additionally, a strong positive correlation was found between SSG FS /min and match FS /min (r = 0.771, p < 0.001). Conversely, significant negative correlations were identified between SSG TD/min and match HSR/min (r = −0.280, p = 0.026), RD/min (r = −0.284, p = 0.024), and FS /min (r = −0.279, p = 0.027), as well as between SSG maximal speed/min and match TD/min (r = −0.260, p = 0.039). These results suggest that while SSGs may partially replicate match demands in specific metrics such as FS, they may not effectively reflect high-speed and RD loads typically observed in competition.
Correlations between SSG sessions and match external load
| SSGs/match | TD/min | PL/min | MaxSpeed/min | HSR/min | ACC/min | DEC/min | RD/min | FS/min |
| TD/min | r = −0.078 | r = 0.300* | r = −0.180 | r = −0.280* | r = −0.183 | r = 0.267* | r = −0.284* | r = −0.279* |
| p = 0.545 | p = 0.017 | p = 0.157 | p = 0.026 | p = 0.151 | p = 0.034 | p = 0.024 | p = 0.027 | |
| PL/min | r = 0.078 | r = 0.184 | r = −0.143 | r = 0.165 | r = −0.096 | r = 0.155 | r = −0.135 | r = 0.310* |
| p = 0.544 | p = 0.149 | p = 0.264 | p = 0.197 | p = 0.454 | p = 0.226 | p = 0.291 | p = 0.014 | |
| MaxSpeed/min | r = −0.260* | r = 0.158 | r = 0.002 | r = −0.129 | r = 0.014 | r = 0.021 | r = −0.031 | r = 0.148 |
| p = 0.039 | p = 0.215 | p = 0.989 | p = 0.314 | p = 0.913 | p = 0.870 | p = 0.807 | p = 0.248 | |
| HSR/min | r = −0.047 | r = 0.415** | r = 0.006 | r = −0.214 | r = 0.014 | r = 0.032 | r = −0.188 | r = −0.102 |
| p = 0.716 | p < 0.001 | p = 0.963 | p = 0.093 | p = 0.910 | p = 0.806 | p = 0.140 | p = 0.424 | |
| ACC/min | r = −0.083 | r = 0.134 | r = −0.097 | r = 0.117 | r = −0.055 | r = 0.077 | r = −0.013 | r = 0.300* |
| p = 0.519 | p = 0.294 | p = 0.448 | p = 0.360 | p = 0.666 | p = 0.548 | p = 0.921 | p = 0.017 | |
| DEC/min | r = 0.164 | r = 0.056 | r = 0.133 | r = −0.023 | r = 0.148 | r = −0.162 | r = 0.006 | r = 0.022 |
| p = 0.198 | p = 0.661 | p = 0.298 | p = 0.859 | p = 0.245 | p = 0.204 | p = 0.965 | p = 0.865 | |
| RD/min | r = −0.075 | r = −0.189 | r = −0.066 | r = −0.212 | r = −0.037 | r = 0.001 | r = 0.129 | r = 0.338* |
| p = 0.560 | p = 0.137 | p = 0.606 | p = 0.095 | p = 0.773 | p = 0.994 | p = 0.311 | p = 0.007 | |
| FS/min. | r = 0.124 | r = −0.146 | r = −0.008 | r = 0.067 | r = −0.002 | r = −0.015 | r = 0.003 | r = 0.771** |
| p = 0.333 | p = 0.255 | p = 0.949 | p = 0.604 | p = 0.986 | p = 0.909 | p = 0.981 | p < 0.001 |
TD: total distance; PL: Playerload; MaxSpeed: maximal speed; HSR: high-speed running; ACC: acceleration; DEC: deceleration; RD: running deviation; SSGs: small-sided games; *: denotes significance at p ≤ 0.05; **: denotes significance at p ≤ 0.001.
The correlations between tactical training sessions’ external load and match external load are presented in Table 4. Significant moderate negative correlations were observed between training TD/min and match HSR/min (r = −0.384, p = 0.008), and between training HSR/min and match RD/min (r = −0.332, p = 0.023). Conversely, a significant moderate positive correlation was found between training HSR/min and match HSR/min (r = 0.436, p = 0.002). Additionally, strong positive correlations were found between training and match values for RD/min (r = 0.678, p < 0.001) and FS /min (r = 0.719, p < 0.001).
Correlations between Tactical sessions and match external load
| Tactical/match | TD/min | PL/min | MaxSpeed/min | HSR/min | ACC/min | DEC/min | RD/min | FS/min |
|---|---|---|---|---|---|---|---|---|
| TD/min | r = −0.012 | r = 0.177 | r = −0.077 | r = −0.384* | r = −0.085 | r = 0.048 | r = −0.060 | r = 0.176 |
| p = 0.937 | p = 0.232 | p = 0.606 | p = 0.008 | p = 0.570 | p = 0.746 | p = 0.686 | p = 0.236 | |
| PL/min | r = −0.043 | r = 0.075 | r = −0.151 | r = −0.248 | r = −0.121 | r = 0.066 | r = −0.058 | r = 0.278 |
| p = 0.773 | p = 0.617 | p = 0.309 | p = 0.092 | p = 0.416 | p = 0.659 | p = 0.698 | p = 0.059 | |
| MaxSpeed/min | r = −0.160 | r = 0.045 | r = −0.136 | r = 0.273 | r = −0.136 | r = 0.137 | r = −0.283 | r = −0.174 |
| p = 0.283 | p = 0.761 | p = 0.359 | p = 0.063 | p = 0.361 | p = 0.359 | p = 0.054 | p = 0.241 | |
| HSR/min | r = −0.125 | r = −0.062 | r = −0.105 | r = 0.436** | r = −0.051 | r = 0.065 | r = −0.332* | r = 0.014 |
| p = 0.402 | p = 0.679 | p = 0.480 | p = 0.002 | p = 0.734 | p = 0.662 | p = 0.023 | p = 0.926 | |
| ACC/min | r = −0.045 | r = −0.103 | r = −0.094 | r = 0.273 | r = −0.097 | r = 0.062 | r = −0.222 | r = 0.082 |
| p = 0.765 | p = 0.493 | p = 0.532 | p = 0.063 | p = 0.518 | p = 0.677 | p = 0.133 | p = 0.585 | |
| DEC/min | r = 0.125 | r = −0.144 | r = −0.076 | r = −0.042 | r = −0.029 | r = 0.115 | r = 0.020 | r = 0.166 |
| p = 0.403 | p = 0.333 | p = 0.610 | p = 0.781 | p = 0.848 | p = 0.439 | p = 0.892 | p = 0.263 | |
| RD/min | r = 0.039 | r = −0.150 | r = 0.015 | r = −0.322* | r = 0.016 | r = −0.091 | r = 0.154 | r = 0.678** |
| p = 0.796 | p = 0.313 | p = 0.922 | p = 0.027 | p = 0.915 | p = 0.541 | p = 0.299 | p < 0.001 | |
| FS/min | r = 0.035 | r = −0.106 | r = −0.059 | r = −0.287 | r = −0.047 | r = 0.003 | r = 0.043 | r = 0.719** |
| p = 0.816 | p = 0.478 | p = 0.694 | p = 0.050 | p = 0.753 | p = 0.987 | p = 0.775 | p < 0.001 |
TD: total distance; PL: Playerload; MaxSpeed: maximal speed; HSR: high-speed running; ACC: acceleration; DEC: deceleration; RD: running deviation; SSGs: small-sided games; *: denotes significance at p ≤ 0.05; **: denotes significance at p ≤ 0.001.
The correlations between combined training sessions’ external load and match external load are presented in Table 5. Significant moderate-to-strong positive correlations were found between training PL/min and match MaxSpeed/min (r = 0.512, p = 0.038), and between training ACCs/min and match PL/min (r = 0.549, p = 0.024). Conversely, significant negative correlations were observed between training PL/min and match MaxSpeed/min (r = −0.495, p = 0.045), training RD/min and match HSR/min (r = −0.603, p = 0.012), and training FS /min and match HSR/min (r = −0.650, p = 0.006).
Correlations between Conditioning + SSGs + Tactical sessions and match external load
| C + SSGs + T/match | TD/min | PL/min | MaxSpeed/min | HSR/min | ACC/min | DEC/min | RD/min | FS/min |
|---|---|---|---|---|---|---|---|---|
| TD/min | r = 0.439 | r = 0.221 | r = −0.091 | r = −0.096 | r = −0.135 | r = 0.279 | r = 0.015 | r = 0.409 |
| p = 0.080 | p = 0.393 | p = 0.730 | p = 0.715 | p = 0.605 | p = 0.276 | p = 0.959 | p = 0.104 | |
| PL/min | r = −0.093 | r = 0.358 | r = −0.495* | r = −0.069 | r = −0.409 | r = 0.407 | r = −0.267 | r = −0.326 |
| p = 0.723 | p = 0.159 | p = 0.045 | p = 0.795 | p = 0.104 | p = 0.106 | p = 0.299 | p = 0.201 | |
| MaxSpeed/min | r = −0.174 | r = 0.512* | r = −0.039 | r = 0.093 | r = 0.059 | r = 0.152 | r = 0.199 | r = −0.218 |
| p = 0.503 | p = 0.038 | p = 0.883 | p = 0.723 | p = 0.824 | p = 0.559 | p = 0.443 | p = 0.399 | |
| HSR/min | r = 0.265 | r = 0.002 | r = −0.088 | r = 0.574* | r = −0.061 | r = 0.211 | r = −0.250 | r = 0.169 |
| p = 0.303 | p = 0.996 | p = 0.737 | p = 0.18 | p = 0.817 | p = 0.415 | p = 0.332 | p = 0.515 | |
| ACC/min | r = −0.145 | r = 0.549* | r = −0.091 | r = 0.162 | r = 0.007 | r = 0.076 | r = 0.162 | r = −0.164 |
| p = 0.579 | p = 0.024 | p = 0.730 | p = 0.534 | p = 0.981 | p = 0.773 | p = 0.534 | p = 0.528 | |
| DEC/min | r = −0.005 | r = −0.311 | r = 0.225 | r = −0.100 | r = 0.169 | r = −0.225 | r = 0.025 | r = −0.037 |
| p = 0.989 | p = 0.223 | p = 0.383 | p = 0.701 | p = 0.515 | p = 0.383 | p = 0.928 | p = 0.891 | |
| RD/min | r = −0.110 | r = −0.135 | r = −0.056 | r = −0.603* | r = −0.120 | r = 0.061 | r = −0.059 | r = −0.108 |
| p = 0.673 | p = 0.605 | p = 0.831 | p = 0.012 | p = 0.646 | p = 0.817 | p = 0.824 | p = 0.680 | |
| FS/min | r = −0.069 | r = −0.228 | r = 0.025 | r = −0.650* | r = −0.054 | r = 0.098 | r = 0.022 | r = 0.201 |
| p = 0.795 | p = 0.377 | p = 0.928 | p = 0.006 | p = 0.839 | p = 0.708 | p = 0.936 | p = 0.438 |
TD: total distance; PL: Playerload; MaxSpeed: maximal speed; HSR: high-speed running; ACC: acceleration; DEC: deceleration; RD: running deviation; SSGs: small-sided games; *: denotes significance at p ≤ 0.05.
A regression analysis revealed several significant predictors of match external load variables (Table 6). Conditioning FS/min significantly predicted match TD/min (β = 0.34, p = 0.02, R 2 = 0.12). SSGs MaxSpeed/min did not significantly predict match TD/min (β = −0.16, p = 0.20). Conditioning HSR/min approached significance for predicting match PL/min (β = 0.29, p = 0.06, R 2 = 0.08). SSGs TD/min (β = 0.27, p = 0.03, R 2 = 0.08) and SSGs HSR/min (β = 0.33, p = 0.008, R 2 = 0.11) significantly predicted match PL/min. The combined variable of C + SSGs + T MaxSpeed/min did not significantly predict match MaxSpeed/min (β = 0.44, p = 0.08, R 2 = 0.19), but C + SSGs + T ACC/min significantly predicted match MaxSpeed/min (β = 0.56, p = 0.02, R 2 = 0.32). The combined variable of C + SSGs + T PL/min significantly predicted match HSR/min (β = 0.42, p = 0.004, R 2 = 0.18). Conditioning FS/min also significantly predicted match HSR/min (β = 0.43, p = 0.003, R 2 = 0.19). SSGs TD/min and tactical TD/min significantly predicted match FS/min (β = 0.52, p < 0.001, R 2 = 0.60), while tactical HSR/min predicted match FS/min (β = 0.84, p < 0.001, R 2 = 0.70). SSGs FS/min showed the highest R 2 value (0.60, p < 0.001) for predicting match FS/min, with additional significant predictors including SSGs TD/min, PL/min, ACC/min, and RD/min. Finally, tactical RD/min and FS/min were also significant predictors, explaining 50% and 70% of the variance, respectively (R 2 = 0.50, p < 0.001, R 2 = 0.70, p < 0.001).
Univariate regression analysis explaining the influence of training external load on match external load
| Predictor variables | Predicted variables | b* | R | R 2 | Adjusted R 2 | F | p |
|---|---|---|---|---|---|---|---|
| Conditioning FS/min | Match TD/Min | 0.34 | 0.34 | 0.12 | 0.10 | 5.63 | 0.02* |
| SSGs MaxSpeed/min | −0.16 | 0.16 | 0.03 | 0.01 | 1.67 | 0.20 | |
| Conditioning HSR/min | Match PL/Min | 0.29 | 0.29 | 0.08 | 0.06 | 3.87 | 0.06 |
| SSGs TD/min | Match PL/Min | 0.27 | 0.27 | 0.08 | 0.06 | 4.92 | 0.03* |
| SSGs HSR/min | Match PL/Min | 0.33 | 0.33 | 0.11 | 0.09 | 7.45 | 0.008* |
| C + SSGs + T MaxSpeed/min | Match PL/Min | 0.44 | 0.44 | 0.19 | 0.14 | 3.51 | 0.08 |
| C + SSGs + T ACC/min | Match PL/Min | 0.56 | 0.56 | 0.32 | 0.27 | 6.98 | 0.02* |
| C + SSGs + T PL/min | Match MaxSpeed/Min | 0.41 | 0.41 | 0.17 | 0.11 | 3.05 | 0.10 |
| Conditioning PL/min | Match HSR/Min | 0.42 | 0.42 | 0.18 | 0.16 | 9.23 | 0.004* |
| Conditioning FS/min | Match HSR/Min | 0.43 | 0.43 | 0.19 | 0.17 | 9.95 | 0.003* |
| SSGs TD/min | Match HSR/Min | 0.23 | 0.23 | 0.05 | 0.04 | 3.38 | 0.07 |
| Tactical TD/min | Match HSR/Min | −0.36 | 0.36 | 0.13 | 0.11 | 6.79 | 0.01* |
| Tactical HSR/min | Match HSR/Min | 0.41 | 0.41 | 0.17 | 0.15 | 9.08 | 0.004* |
| Tatical RD/min | Match HSR/Min | −0.31 | 0.31 | 0.10 | 0.08 | 4.78 | 0.03* |
| C + SSGs + T HSR/min | Match HSR/Min | 0.51 | 0.51 | 0.26 | 0.21 | 5.19 | 0.03* |
| C + SSGs + T/min. RD/min | Match HSR/Min | −0.66 | 0.66 | 0.44 | 0.41 | 11.87 | 0.004* |
| C + SSGs + T FS/min | Match HSR/Min | −0.60 | 0.60 | 0.36 | 0.32 | 8.49 | 0.01* |
| Conditioning PL/min | Match ACC/Min | 0.31 | 0.31 | 0.10 | 0.10 | 4.58 | 0.3* |
| Conditioning FS/min | Match ACC/Min | 0.25 | 0.25 | 0.06 | 0.04 | 2.86 | 0.09 |
| Conditioning PL/min | Match DEC/Min | −0.27 | 0.27 | 0.07 | 0.05 | 3.32 | 0.07 |
| SSGs TD/min | Match DEC/Min | 0.08 | 0.08 | 0.00 | −0.01 | 0.38 | 0.54 |
| Conditioning FS/min | Match RD/Min | 0.11 | 0.11 | 0.01 | −0.01 | 0.52 | 0.47 |
| SSGs TD/min | Match RD/Min | −0.12 | 0.12 | 0.01 | −0.00 | 0.88 | 0.35 |
| Tatical HSR/min | Match RD/Min | −0.28 | 0.28 | 0.08 | 0.06 | 3.73 | 0.06 |
| Conditioning FS/min | Match FS/min | 0.52 | 0.52 | 0.27 | 0.25 | 15.69 | <0.001** |
| SSGs TD/min | Match FS/min | −0.33 | 0.33 | 0.11 | 0.09 | 7.35 | 0.009* |
| SSGs PL/min | Match FS/min | 0.26 | 0.26 | 0.07 | 0.05 | 4.51 | 0.03* |
| SSGs ACC/min | Match FS/min | 0.33 | 0.33 | 0.11 | 0.09 | 7.37 | 0.009* |
| SSGs RD/min | Match FS/min | 0.40 | 0.40 | 0.16 | 0.15 | 11.88 | 0.001* |
| SSGs FS/min | Match FS/min | 0.77 | 0.77 | 0.60 | 0.60 | 90.40 | <0.001** |
| Tatical RD/min | Match FS/min | 0.70 | 0.70 | 0.50 | 0.49 | 44.27 | <0.001** |
| Tatical FS/min | Match FS/min | 0.84 | 0.84 | 0.70 | 0.70 | 105.66 | <0.001** |
FS: footstrikes; TD: total distance; MaxSpeed: maximum speed; HSR: high-speed running; PL: playerload; RD: running deviation; ACC: acceleration; DEC: deceleration; SSGs: small-sided games; C + SSGs + T: conditioning + small-sided games + tatical. *: denotes significance at p ≤ 0.05; **: denotes significance at p ≤ 0.001.
The present study aimed to analyze the influence of the external load of SSGs, tactical, conditioning, and combined sessions (SSGs + T, + C), on the subsequent match demands. Conditioning sessions showed positive correlations with PL, HSR, and FS, and negative correlations with RD. FS predicted a match TD. SSGs were positively correlated with PL, HSR, and FS, and negatively with MaxSpeed, RD, and TD. SSGs TD and HSR predicted PL. C + SSGs + T ACC predicted MaxSpeed, and C + SSGs + T PL and conditioning FS predicted HSR. SSGs TD and tactical TD predicted FS, with SSGs FS showing the strongest predictive value. Tactical RD and FS predicted RD and FS, explaining 50 and 70% of the variance, respectively.
SSGs training session type showed several significant correlations with match external load metrics, notably with PL/min, HSR/min, and FS/min. The positive correlations between SSGs and PL, HSR, and FS highlight the relevance of SSGs in replicating match demands, especially high-intensity movements such as ACCs and DECs. SSGs have been well-documented in previous literature as an effective training method for simulating match-specific physical demands (Castillo et al., 2021). Our results further support this by showing that SSGs not only increase external load but also increase FS. These findings are consistent with previous studies reporting that SSGs elicit high external and internal loads comparable to match play, particularly through frequent changes in direction, ACCs, and DECs (Savolainen et al., 2023). The higher FS/min observed may reflect the repeated multidirectional actions and short bursts of movement inherent in SSGs (Xu et al., 2021). This increased neuromuscular and mechanical load has been associated with greater PL (Collins et al., 2024). Moreover, the elevated PL/min suggests that SSGs stimulate match-relevant intensity. Moreover, SSGs TD and HSR significantly predicted match PL, with the stronger predictive power of SSGs HSR supporting previous observations that well-designed SSGs elicit HSR stimuli transferable to match play (Clemente et al., 2021). In contrast, SSGs MaxSpeed did not predict match TD, likely due to the spatial constraints in SSGs limiting peak speed attainment (Makar et al., 2023).
Interestingly, tactical sessions did not demonstrate strong correlations with external load measures, particularly HSR, and DECs. This aligns with previous studies showing that tactical exercises typically impose lower physical loads compared to formats such as SSGs or HIIT, due to their emphasis on positioning, game strategy, and decision-making rather than sustained high-intensity efforts (Clemente et al., 2022). Nevertheless, the observed moderate positive correlation between tactical HSR/min and match HSR/min suggests that these sessions involve significant movement demands. Additionally, a significant negative correlation between tactical RD/min and match HSR/min was found in the present study. This further supports the idea that tactical sessions contribute to spatial adjustments on the pitch (Batista et al., 2019). Tactical TD and HSR were significant predictors of match FS. Thus, tactical RD and FS accounted for 50% and 70% of the variance in match RD and FS, respectively, indicating that the movement patterns observed during tactical sessions are strongly related to in-game performance. This emphasizes the potential of tactical conditioning to mimic key aspects of match play, especially in terms of spatial adjustments and dynamic movements. These results support the integration of tactical training to improve match-specific performance, particularly in terms of FS frequency and RDs. Although these sessions are less physically intense than match play, they may still induce neuromuscular load, emphasizing the need for their strategic integration into the periodized training plan.
Conditioning sessions showed mixed associations with match external load, with significant correlations observed. This suggests that general conditioning may contribute to foundational aerobic fitness, supporting overall performance, particularly for players with lower high-intensity demands (Buchheit & Laursen, 2013). However, the weak relationships with high-intensity metrics such as maximal speed and HSR do not reflect findings in previous research indicating that traditional conditioning formats are superior in replicating match-specific intensities (Clemente, 2020; García-Ramos et al., 2018). Conditioning FS significantly predicted match TD, indicating that high-frequency locomotor efforts in conditioning reflect continuous match demands, which aligns with previous findings (Gualtieri et al., 2023). Conditioning FS is also strongly predicted to match HSR. Although conditioning HSR only had a significant correlation with match PL, it suggests that incorporating high-speed conditioning exercises may further increase load specificity.
Finally, the combined C + SSGs + T ACC significantly predicted match MaxSpeed, showing the relevance of ACC measure as a precursor to maximal sprint performance (Young et al., 2018), while both C + SSGs + T PL strongly predicted match HSR. Integrating ACC and PL-focused training may increase the athletes’ readiness for the speed and intensity demands of match play. ACC training, which improves explosive ACC, helps athletes generate the quick bursts of speed needed for maximal sprints during games (Haugen et al., 2014; Little & Williams, 2005). On the other hand, PL-focused training, which targets overall physical load and HSR, prepares athletes to sustain intense efforts over longer periods (Beato et al., 2021). Combined, these training types improve both explosive power and endurance, enabling athletes to perform effectively in dynamic match scenarios, reduce injury risks, and better manage the physical demands of competitive play.
Several limitations should be considered. First, the observational design limits the ability to infer causality, and the results may be influenced by unmeasured factors such as the individual players’ prior conditioning and external environmental conditions during the study period. Second, the observational design and relatively small sample size of 24 female players from a single national team significantly limit the generalizability of our findings. The homogeneous sample (national-level players with similar training backgrounds) restricts extrapolation to other competitive levels (e.g., amateur, youth, or professional club settings) or male populations, where training responses and match demands may differ substantially. The single-team design also prevents accounting for inter-team variability in coaching philosophies, tactical systems, and training methodologies, which could influence the training-to-match load relationships observed. Additionally, the observational nature precludes causal inference, as we cannot control confounding variables such as individual player conditioning, recovery status, or environmental factors during the study period. Further studies involving larger sample sizes across different teams, and competitive levels would provide a broader understanding of the observed patterns. Additionally, future research could incorporate a longitudinal design to assess how changes in external load across the season influence performance outcomes. Another limitation is the reliance on GPS technology for tracking external load measures, which, while reliable, may not capture all determinant aspects of player movements, such as technical actions or cognitive demands during training and match play. Future studies could integrate biomechanical and physiological assessments for a more comprehensive understanding of the factors influencing match performance.
This study provides evidence-based guidance for coaches designing training sessions to optimally prepare elite female soccer players for match demands. SSGs effectively mirror key match demands such as PL and HSR, with strong relationships to match FS indicating that SSGs replicate both metabolic and neuromuscular competition demands. However, conditioning sessions are particularly effective for developing ACCs and DECs – critical components of match performance that may be underemphasized in SSG-only approaches, with high-frequency FS during conditioning translating to greater overall match distance coverage. While tactical sessions impose lower physical loads than SSGs or conditioning, they strongly influence match-specific movement patterns including RDs and FS, emphasizing their role in preparing players for positional and spatial demands of competition. Combined sessions integrating conditioning, SSGs, and tactical elements contribute uniquely to MaxSpeed development and HSR capacity, suggesting that periodized integration of multiple training formats optimizes comprehensive match readiness. In summary, coaches should not rely exclusively on any single training format, as a balanced weekly microcycle incorporating conditioning for explosive capacity, SSGs for match-specific intensity, tactical work for movement patterns, and combined formats for comprehensive preparation appears essential for preparing elite female soccer players for the multifaceted demands of competitive match play.
Rui Miguel Silva is a research member of the Sport Physical Activity and Health Research & Innovation Center (SPRINT) which is funded by the Portuguese Foundation for Science and Technology (FCT) under the identifier UID/06185/2025, DOI: 10.54499/UID/06185/2025.
FCT – Fundação para a Ciência e a Tecnologia – Foundation for Science and Technology, I.P. (Portugal), within the scope of SPRINT – Sport Physical Activity and Health Research & Innovation Center [UID/6185/2023] with the DOI: 10.54499/UID/PRR/06185/2025.
Conceptualization, R.M.S.; methodology, R.M.S. and M.S.; formal analysis, R.M.S.; investigation, R.M.S., M.S., Z.I. and D.M.; writing – original draft preparation, R.M.S.; writing – review and editing, R.M.S., M.S., Z.I. and D.M. All authors read and approved the final version of the manuscript.
Authors state no conflict of interest.