Different Aspects of Physical Load in Small-Sided Field Hockey Games

Abstract

Wilmes, E, de Ruiter, CJ, van Leeuwen, RR, Banning, LF, van der Laan, D, and Savelsbergh, GJP. Different aspects of physical load in small-sided field hockey games. J Strength Cond Res 38(2): e56–e61, 2024—Running volumes and acceleration/deceleration load are known to vary with different formats of small-sided games (SSGs) in field hockey. However, little is known about other aspects of the physical load. Therefore, the aim of this study was to gain a more thorough understanding of the total physical load in field hockey SSGs. To that end, 2 different SSGs (small: 5 vs. 5, ∼100 m² per player; large: 9 vs. 9, ∼200 m² per player) were performed by 16 female elite field hockey athletes. A range of external physical load metrics was obtained using a global navigational satellite system and 3 wearable inertial measurement units on the thighs and pelvis. These metrics included distances covered in different velocity ranges (walk, jog, run, and sprint), mean absolute acceleration/deceleration, Hip Load, and time spent in several physically demanding body postures. The effects of SSG format on these external physical load metrics were assessed using linear mixed models (p < 0.05). Running volumes in various speed ranges were higher for the large SSG. By contrast, mean absolute acceleration/deceleration and time spent in several demanding body postures were higher for the small SSG. This study shows that changing the SSG format affects different aspects of physical load differently.

Introduction

Field hockey is a team sport that is associated with a high physical load (18,20). This high physical load must be balanced with sufficient recovery to improve performance and reduce injury risk (26). To achieve an adequate balance between the physical load and recovery, coaches manipulate the physical load during training. One way to do so is by varying pitch size and the number of players on the pitch during training matches (10,13,24,25). Therefore, similar to other field-based team sports (6,9,15,23), small-sided games (SSGs) in field hockey training are often used not only to manipulate technical and tactical demands but also to control the physical load of the game. Although a considerable amount of research has been performed on the running load of SSGs (e.g., total distance, distances above predefined velocity thresholds, and accelerations/decelerations) (10,13,20), little is known about other aspects of the physical load, such as the time spent in physically demanding body postures and joint angular accelerations.

When assessing physical load, a distinction can be made between internal and external load, referring to whether the measurable aspect of the load occurs within or outside of the athletes' body (16). The external load to which an athlete is subjected brings about internal responses to withstand the demands elicited by the external load. These internal responses are generally referred to as internal load and can be categorized under physiological and biomechanical loads (26). Physiological load is defined as the metabolic and cardiovascular aspect of the internal load (e.g., heart rate and energy consumption), whereas biomechanical load is defined as mechanical stresses and strains on musculoskeletal tissues (e.g., muscle strain and tendon force). Clearly, physiological and biomechanical aspects of the internal load are interrelated (26). Metabolic and cardiovascular processes are necessary to generate muscle forces. However, because musculoskeletal stresses and strains are determined by internally generated forces and external forces, the type of exercise or movement determines the relative magnitudes of physiological and biomechanical load. For example, running and cycling at certain speeds may induce similar physiological load (e.g., heart rate). Yet, because of impacts with the ground, biomechanical load (e.g., peak muscle forces) is higher during running.

Ultimately, training outcome results from the balance between internal load and recovery because the human body adapts to training stimuli (16). It should be noted, however, that recovery rates depend on the type of stimulus (26). Therefore, to increase control on training outcome, it would be desirable to measure both the physiological and biomechanical aspects of the internal load of an athlete. However, despite this desire, measurements of internal load are often impossible (e.g., muscle force (27)), not feasible in practice (e.g., energy consumption), or not valid for intermittent high-intensity team sports (e.g., heart rate (16)). Accordingly, much of practice is limited to the use of external load metrics, which are aimed to reflect the internal load and may serve as proxies. Yet, external load metrics can be designed to mostly reflect internal physiological load (e.g., covered distance) or to mostly reflect internal biomechanical load (e.g., decelerations).

Currently, most external training load metrics used in field hockey are based on global navigation satellite systems (GNSS), which are worn between the scapulae and measure whole-body location on the field. Although such systems provide useful information regarding external training load (i.e., total distance covered, sprint distance, accelerations, and decelerations) (10,13,20), most of the external load metrics obtained from these systems are aimed to reflect whole-body physical load. A more valid and detailed reflection of the physical load of field hockey may be obtained when lower limb kinematics are measured because most load and injuries in field hockey concern the muscles around the hips (2,3,29,30). Recent studies have used wearable inertial measurement units (IMUs) to obtain such metrics (3,31). These metrics include the time spent in different demanding body postures and Hip Load, which is the squared magnitude of hip angular acceleration divided by a scale factor. Because hip angular accelerations are required for whole-body accelerations, decelerations, and running at (nearly) constant speeds, Hip Load may be a better reflection of the total running load compared with the distances covered in different velocity bands or whole-body acceleration/deceleration measures alone (29). These metrics make it possible to gain a more thorough understanding of the physical load of SSGs in field hockey in addition to the running load.

Compared with full-sized matches, SSGs have been shown to result in a lower maximal 5-minute mean player velocity, more time spent at a low velocity (<1 m·s⁻¹), and less time spent at a moderate (1–5 m·s⁻¹) to high velocity (>5 m·s⁻¹) (10,13), indicating higher volumes of running at higher speeds in full-sized matches. Conversely, a higher maximal 5-minute mean absolute acceleration/deceleration has been found in SSGs (10), indicating a higher acceleration/deceleration load in SSGs. In addition, in field hockey specifically, when players are involved with the ball, they must reach to the ground with their stick, leading to a posture with a forward flexed trunk and at least 1 flexed thigh. This posture is associated with a relatively high hamstring and gluteal muscle activity (11,22), suggesting a high load on these muscles. Furthermore, field hockey players spent a considerable amount of time with their trunk flexed forward (28), leading to greater hip and back extension moments compared with standing or running upright. When the number of players on the pitch is reduced, players are more often involved with the ball. Therefore, it is expected that players spent more time in these physically demanding body postures when the number of players on the pitch is reduced.

The aim of this study was to gain a more thorough understanding of the different aspects of physical load of 2 different SSG formats in field hockey. Accordingly, small SSG (5 vs. 5, ∼100 m² per player) and large SSG (9 vs. 9, ∼200 m² per player) training matches were executed by a group of elite female field hockey players. A range of different kinematic variables were obtained using a GNSS and wearable IMUs, including distances covered in different velocity bands, mean absolute acceleration/deceleration, the time spent in demanding body postures, and Hip Load. A higher volume of running distance was expected for all velocity ranges for the large SSG, but a higher acceleration/deceleration load and more time spent in demanding body postures were anticipated for the small SSG.

Methods

Experimental Approach to the Problem

A prospective observational research design was adopted to examine different aspects of the physical load of 2 SSG formats in female elite field hockey players. Experimental data were collected during 2 regular training sessions in the same week that were 2 days apart. The testing procedures were approved by the Local Ethics Committee of the Vrije Universiteit Amsterdam (VCWE-2019-070R1) and were in accordance with the Declaration of Helsinki.

Subjects

Sixteen elite female field hockey players of the Dutch national team under 21 participated in the study (mean ± SD, age, 19.3 ± 0.9 years; body mass, 63.7 ± 4.4 kg; height, 172.0 ± 5.4 cm; and playing experience, 12.9 ± 1.4 years). Goalkeepers were excluded from the experiments. All subjects were free of injury at the time of testing, were informed about all testing procedures before the experiments, and signed informed consent.

Procedures

Before the start of each training session, the subjects were equipped with a GNSS sensor and 3 wearable IMUs. The IMUs were fixated to each thigh and the pelvis (further details will be given below). Subjects completed a 25-minute warm-up procedure that was designed by the teams' physiotherapist and included running, a passing and shooting exercise, and stretching. Thereafter, the following 2 types of SSGs were played; first, a 5 vs. 5 small SSG on a pitch-sized 25 × 40 m translating to ∼100 m² per player; and later, a large SSG on a pitch-sized 55 × 65 m translating to ∼200 m² per player, which is roughly equal to the player density in a full-sized match (11 vs. 11 on a pitch-sized 91.4 × 55, 228.5 m² per player). Both SSGs were played for approximately 20 minutes and included regular goals and goalkeepers.

Equipment

Subjects were equipped with a GNSS tracker (JOHAN V4, JOHAN Sports, Delft, the Netherlands) worn between the scapulae, which sampled the player position at a 10 Hz sampling frequency. In addition, the subjects were equipped with 3 wearable 9-DOF IMUs (MPU-9150, InvenSense, San Jose, CA) 1 on the pelvis and 1 on each thigh, which were embedded in a protective casing together with a battery and an SD card with a total weight of 11 grams. Before the IMUs were attached to the player, they were time synchronized by introducing a mechanical synchronization peak according to de Ruiter et al. (7). Thereafter, they were placed in small pockets that were sewn into short tights that the players wore underneath their normal hockey outfit (Figure 1). Each IMU measured 3D acceleration, 3D angular velocity, and 3D magnetic field strength of the body segment it was attached to at a 500 Hz sampling frequency. Sensor-to-body calibrations were performed according to Wilmes et al. (30), and 3D body segment orientations were tracked continuously using a Madgwick gradient descent orientation filter with a filter gain β of 0.043 (19). Based on the 3D body segment orientations, segment angles with the vertical in the body's sagittal plane were determined (Figure 2).

Figure 1.

Short tights with inertial measurement units. Small pockets were sewn into the tights at the pelvis and about halfway the thighs on the lateral side. Inertial measurement units were inserted into these pockets. This short tight was worn underneath the hockey shorts (not on the picture).

Figure 2.

Schematic representation of measurement setup and body segment angles. Subjects wore a global navigation satellite system (GNSS) unit on their back between the scapulae and inertial measurement units (IMUs) on the pelvis and each thigh. Segment angles were defined in the body's sagittal plane between the segment and the vertical. Body segment angles are defined as positive when the body segment is in front of the hips and as negative when the body segment is behind the hips. When the player stands completely upright, all body segment angles are 0°.

Statistical Analyses

Cumulative external physical load metrics collected for the small and large SSGs were all expressed per minute of playing time or as the mean value over the total playing time. The moments when the coach would stop play to give feedback were cut out of the data because these were not part of the SSGs and, if included, would give a distorted view on the true load during play.

The external load metrics based on the GNSS data were the following: total distance (TD; expressed in m·min⁻¹), jogging distance (JD; distance covered between 7 and 14 km·h⁻¹, expressed in m·min⁻¹), running distance (RD; distance covered between 14 and 20 km·h⁻¹, expressed as m·min⁻¹), sprint distance (SD; distance covered above 20 km·h⁻¹, expressed as m·min⁻¹), and the mean absolute acceleration and deceleration (Mean_AccDec).

In addition, external load metrics based on the IMU data were as follows: (a) Time spent in a lunge position with a forward flexed trunk (Figure 1). The boundaries for this specific body posture were defined as a minimum thigh angle of 30° (front leg) and a concomitant minimum pelvis angle of 30°. This was calculated separately for the left leg being the front leg (T_lunge left) and the right leg being the front leg (T_lunge right). (b) The time spent in 2 different ranges of thigh angles for the left and right leg. These ranges included a relatively shallow squat position with thigh angles between 30° and 60° (T_{thigh, 30–60}) and a deeper squatted posture with thigh angles of 60° and above (T_{thigh, >60}). (c) The time spent in different ranges of the pelvic tilt angle. The ranges included the following pelvis angles from upright to a more anteriorly tilted pelvis: 10° to 20° (T_{pelvis, 10–20}), 20° to 30° (T_{pelvis, 20–30}), and 30° and above (T_{pelvis, >30}). (d) The recently developed Hip Load (HL; equation 1) metric, which has shown to be a valid and reliable metric in football (3,29). Hip Load is the squared magnitude of hip angular acceleration divided by a scale factor to improve readability and is expressed in arbitrary units. The HL was also calculated for both legs.

$$Hip\hspace{0.17em}Load={\frac{\left|{\alpha }_{hip}\right|}{{10}^8}}^2$$ (1)

Statistics

The effect of the type of SSG on each external load metric was determined using separate linear mixed models using the “lme4” package in R (version 4.2.2) (4,21). The SSG type was entered as a fixed factor, and all models included a random intercept for subjects to account for interindividual differences. The training session (both SSG formats were played during 2 different training sessions) was also included as a random intercept if the inclusion of this factor would lead to a model improvement based on the Akaike information criterion. The inclusion of this factor was to take potential differences between the training sessions into account. Effect sizes were calculated as Cohen's d and were interpreted as small (d = 0.2), medium (d = 0.5), and large (d > 0.8). Effects were deemed significant with p < 0.05.

Results

The results of the external load metrics and linear mixed models are shown in Table 1. In addition, the individual values and means per SSG type are shown in Figure 3. Not all players were present at both training sessions because of various reasons (illness, injury, or school). Furthermore, we encountered some technical issues, such as sensor failures, resulting in the loss of data. Therefore, the number of observations on which the mixed models were based is also shown in Table 1. Large effects of the SSG format can be observed for distances covered, whereby significantly more distance was covered during the large SSGs, except for sprint distance. By contrast, opposite large effects with higher values for the small SSG were observed for the mean absolute acceleration and deceleration, the time spent in all reported ranges of thigh angles, and the time spent with the pelvis angle between 20° and 30°.

Table 1.

Results of external load metrics and effects of linear mixed models.^*

	5 vs. 5 SSG	9 vs. 9 SSG	Effect SSG format	Model
	Mean ± SD	Mean ± SD	Est.	95% CI	Cohen's d	p	R 2	No. of obs.
GNSS-based metrics
Total distance (m·min−1)†	109.5 ± 10.4	121.7 ± 10.9	11.8	8.4 to 15.2	2.49	<0.01	0.82	47
Jog distance (m·min−1)†	53.3 ± 8.3	61.8 ± 8.0	8.2	5.4 to 10.9	2.12	<0.01	0.78	47
Run distance (m·min−1)†	17.5 ± 7.1	21.2 ± 7.6	3.3	1.0 to 5.7	1.05	<0.01	0.77	47
Sprint distance (m·min−1)†	1.4 ± 1.4	2.3 ± 1.9	0.8	−0.0 to 1.7	0.69	0.055	0.40	47
MeanAccDec (m·s−2)†	0.56 ± 0.14	0.49 ± 0.08	−0.08	−0.13 to −0.03	1.13	<0.01	0.70	44
IMU-based metrics
Tlunge left (s·min−1)	0.15 ± 0.19	0.19 ± 0.30	0.02	−0.09 to 0.13	0.16	0.693	0.57	40
Tlunge right (s·min−1)	0.15 ± 0.19	0.18 ± 0.24	0.02	−0.07 to 0.11	0.17	0.663	0.60	40
Tthigh, 30–60 left (s·min−1)	12.8 ± 5.0	10.3 ± 3.7	−2.66	−4.22 to −1.09	1.32	<0.01	0.73	45
Tthigh, 30–60 right (s·min−1)	12.5 ± 4.9	11.3 ± 3.4	−1.61	−2.97 to −0.25	0.89	0.021	0.79	45
Tthigh, >60 left (s·min−1)	0.32 ± 0.27	0.19 ± 0.18	−0.15	−0.26 to −0.04	0.98	0.010	0.44	45
Tthigh, >60 right (s·min−1)	0.37 ± 0.34	0.22 ± 0.17	−0.15	−0.28 to −0.02	0.85	0.023	0.44	45
Tpelvis, 10–20 (s·min−1)	8.57 ± 5.29	6.76 ± 5.33	−1.92	−4.13 to 0.28	0.65	0.085	0.55	47
Tpelvis, 20–30 (s·min−1)	2.36 ± 2.39	1.51 ± 1.55	−0.93	−1.59 to −0.26	1.04	<0.01	0.74	47
Tpelvis, >30 (s·min−1)	0.55 ± 0.65	0.52 ± 0.62	−0.05	−0.25 to 0.16	0.17	0.641	0.73	47
HL left (AU·min−1)†	101.5 ± 19.6	99.1 ± 23.4	−4.0	−10.3 to 2.4	0.52	0.217	0.78	40
HL right (AU·min−1)†	107.3 ± 17.5	107.3 ± 16.7	−2.9	−8.3 to 2.5	0.44	0.284	0.88	40

^*The external load metrics during the SSG formats are shown as mean ± SD. In addition, the fixed effects of each model are shown as the effect estimate (est.), 95% confidence interval (95% CI) of the effect, effect size (Cohen's d), and p value.

^†Denotes the models that included a random intercept for the training session. The 5 vs. 5 small-sided game (SSG) was modeled as the intercept. Therefore, a positive effect of the SSG format means that the load metric was higher for the large SSG, whereas a negative effect means that the load metric was lower for the large SSG. Significant effects are in bold face. Furthermore, the explained variance (R²) and the number of observations (No. of obs.) are shown for each model.

Figure 3.

Individual and mean values of load metrics small-sided games (SSGs). Significant differences between the small SSGs (5 vs. 5) and the large SSGs (9 vs. 9) are denoted by an asterisk (*).

Discussion

The aim of this study was to gain a more thorough understanding of the different aspects of the physical load of SSG formats in field hockey. Higher running volumes were expected for the large SSGs, but a higher mean absolute acceleration/deceleration and more time spent in physically demanding body postures were anticipated for the small SSG. Distances covered in all quantified velocity ranges were significantly higher for the large SSGs compared with the small SSGs, except sprint distance. However, mean acceleration/deceleration and the time spent in several demanding body postures were significantly higher for the small SSGs. These results indicate that the physical load differs between different SSG formats.

In accordance with the present results, a previous study in elite female field hockey players has demonstrated that the covered distance is greater on a larger pitch with a lower player density (i.e., more available space per player), whereas mean absolute acceleration/deceleration is higher on a smaller pitch with a higher player density (10). More available space allows players to cover greater distances without getting into a duel with another player. In addition, a larger pitch with a lower player density means more time and space available to accelerate to higher running speeds, which is not only reflected by a greater total distance but also by more distance covered at higher running speeds. By contrast, by decreasing the pitch size and space per player, less time and space is available to accelerate to these higher running speeds. Furthermore, less space means that it is harder to obtain a free line of passing, probably necessitating more accelerations and decelerations to obtain the space to receive a ball. Similar results have been found in elite football players (14), where larger pitch sizes and more available space per player also lead to more distance covered but to fewer accelerations and decelerations. In general, therefore, it seems that increasing the pitch size and available space in SSGs leads to greater volumes of running at different speeds, whereas the opposite is true for the acceleration/deceleration load.

Another important finding was that players spent more time in most of the physically demanding body postures during the small SSGs. Unexpectedly and inconsistent between subjects, similar time was spent in a lunge position with a forward flexed trunk during both SSG formats. Presumably, during the large SSGs, players pass over greater distances. With such passes, more kinetic energy must be given to the ball to cover the distance at a high speed. Consequently, players may make a bigger forward step to prolong the push of the ball, leading to concomitant large pelvis and thigh angles. On the other hand, the small SSG may involve more defending actions during which this posture may be adopted more often but for shorter periods of time. Moreover, more time was spent with thigh angles above 30° during the small SSGs and with a pelvis angle greater than 10°. Such body postures are likely to be adopted when players are involved with the ball because they need to reach to the ground with their stick (31). Therefore, these results suggest that limiting the number of players on the pitch does not only lead to more technical actions performed per player (24,25) but also lead to more time spent in physically demanding body postures.

Interestingly, although the same effect of the SSG type on the time spent in demanding body postures was found for most players, absolute differences between the players were relatively large (Figure 3). These differences could be attributed to one or several of the following reasons in an arbitrary order. First, players may use different strategies to reach the ground with the stick. Either the hips can be flexed or the knees can be bent to be able to play the ball. For example, an attacker running with the ball may use a hip-dominant strategy, so a high running velocity can be reached while driving the ball. On the other hand, when a high running velocity is not required, players may opt for a knee-dominant strategy. Second, players may have different anthropometrics, which could make either a hip-dominant strategy or a knee dominant strategy more comfortable for a particular player. For example, it has been shown that people with a shorter thigh relative to their shank are able to reach a deeper squat (8). Therefore, it may be possible that players with relatively short thighs would opt for a more knee-dominant strategy to reach for the ball. Third, relatively large offsets in body frame definitions can be found between IMU-based sensor-to-body calibrations and sensor-to-body calibrations based on anatomical landmarks (17). Although these offsets are presumably consistent within a player, it may partially explain the large differences between players. Therefore, differences in these training load measures between training sessions should be interpreted within the individual. It may be worthwhile to construct individualized boundaries of body segment angles associated with specific hockey actions.

In contrast to our expectations, no significant differences in HL were found between the small and large SSGs. HL is aimed to quantify the total external load on the muscles around the hip, whereby the joints' angular accelerations are squared to give high-intensity accelerations more weight in the load estimate (3,29). These high angular accelerations can occur during different types of movements. For example, sprint running would lead to large hip angular accelerations, but short explosive whole-body accelerations and decelerations can also be the source of high joint angular accelerations. Because distances covered and the acceleration/deceleration load were clearly different between the SSG types, the similar HL values may originate from different movement patterns between the SSGs. This may indicate that the total external load on the muscles around the hip may have been similar between both SSG formats. Therefore, the inclusion of HL in training load monitoring may lead to a more complete view of the total load on the hip muscles compared with monitoring running distances and accelerations/decelerations alone.

This study was not without limitations. First, because we performed the experiments during regular training sessions, we had no full control of the experimental design. Therefore, the small SSGs were played first in both training sessions, whereas counterbalancing the order of SSG types between the 2 training sessions would have been better. However, it should be noted that these players were highly trained female athletes; thus, any order effect was probably small. Second, the pitch area and player density were simultaneously manipulated, which makes it impossible to infer the effects of these manipulations in isolation. Future research should aim to manipulate player density and pitch size separately, such that inferences can be made on the individual effects of these SSG manipulations on the physical load. Third, in the analysis, we did not distinguish between playing positions. When there are only a few players participating in an SSG, all players function as an attacker and defender at the same time, whereas these roles may be more clearly defined with more players on the field. As a result, the effects of different SSG types may be different for players playing at different positions. In addition, we included only an elite female population in this study. Further work is required to establish whether the same effects occur in different athlete populations. Fourth, although test-retest reliability has shown to be good for HL (29), this remains to be seen for the other IMU-based variables tested in this study. Finally, we chose the mean absolute acceleration/deceleration as a metric for the acceleration load. Although the measurement of accelerations and decelerations within specific ranges may provide more detailed information regarding this load, limited validity regarding the measurement of instantaneous accelerations with the GNSS has repeatedly been reported (1,5,12), especially around changes of direction (12), which are common during SSGs.

Notwithstanding these limitations, this study confirmed the existing literature in showing higher volumes of running during larger game formats with more players on the pitch and more available space per player, whereas the acceleration/deceleration load appeared to be higher for a small SSG format with less players on the pitch and less available space per player. In addition, this study adds new information regarding the physical load differences between the 2 SSG formats. More specifically, it was shown that more time was spent in physically demanding body postures during the small SSG format, whereas HL was similar between the SSG types. These results clearly show that the physical load was different in the 5 vs. 5 SSG format compared with the 9 vs. 9 SSG format.

Practical Applications

The findings of this study have an important implication for future practice. This study has shown that practitioners should be aware that changes in the SSG format not only affect running volumes but also the acceleration/deceleration load and the time spent in physically demanding body postures. For example, larger SSGs can be embedded in a training session when the emphasis is on running volume, whereas smaller SSGs can be used to emphasize the acceleration/deceleration load and physical load originating from the time spent in physically demanding body postures. Practitioners can use this information to adequately periodize the design of their training programs.