TY - JOUR AU1 - Hoppe, Matthias W. AU2 - Hotfiel, Thilo AU3 - Stückradt, Alexandra AU4 - Grim, Casper AU5 - Ueberschär, Olaf AU6 - Freiwald, Jürgen AU7 - Baumgart, Christian AB - Introduction Tennis match play involves short repeated high-intensity activities over an unpredictable time. The rules of the International Tennis Federation (ITF) mandate that high-intensity periods are separated by recovery intervals of predefined durations [1]. Over the past 20–30 years, tennis has evolved into a physically demanding sport in all age groups and both genders [2]. Keeping pace with this progress requires specific training drills for which knowledge of match play data is essential [3]. In tennis, match play data have been separated into data describing the activity profiles, technical-tactical actions, mechanical power outputs, and physiological responses of the players [4, 5]. Mainly in males, many studies summarized in reviews have investigated the activity profiles and physiological responses [1, 2], whereas only few studies have examined the mechanical power outputs and technical-tactical behaviors [6–12]. This lack of research is surprising, because both aspects strongly determine the multifactorial tennis performance [13, 14]. From a motor control perspective, superior to the technical-tactical behaviors of tennis players are their playing strategies. To explain, technical-tactical behaviors are variable during play and do take the behavior of opponents into account. In contrast, playing strategies are often predefined, self-referential, and thus fixed over a certain or the entire playing time [15]. Although playing strategies are barely defined and investigated in tennis [8, 16], two contrary strategies can be observed in practice: The number of winners is only at top playing levels comparable to that of errors. Thus, a common strategy to succeed is to reduce the own errors by a passive play from the baseline. The contrary strategy is to dominate the rallies by an active play involving powerful topspin strokes at sharp angles across the full court. The goal of this strategy is to force the opponent to errors or to directly win the points by oneself [4, 17]. In 2003, a framework to describe the physical demands of players was introduced. According to external-mechanical or internal-physiological measurable data, it was suggested to differentiate between external and internal loads, respectively [18]. In tennis, external loads have been mainly assessed by positioning systems and inertial measurement units. These technologies allow to quantify distances covered and various speed, acceleration, and deceleration measures [7, 19, 20]. For internal loads, short-range telemetry, portable respiratory gas analyzers, capillary blood samples, and subjective scales have often been used. By these procedures, heart rates, oxygen uptakes, energy expenditures, blood lactate concentrations, and ratings of perceived efforts can be determined [1, 21]. Importantly, external and internal loads are connected by the concept of mechanical efficiency. This means that physical training has either the goal to improve the mechanical power output at a given physiological response or to decrease the physiological response at a given mechanical power output [22]. Consequently, in tennis, only knowledge of both external and internal loads will lead to ideal designed training drills [23]. In tennis, one previous study examined the effects of passive and active playing strategies on only internal loads in male players [16]. The study shows that a passive strategy leads to a higher oxygen uptake, ventilation, heart rate, and blood lactate concentration (all p<0.001) than an active strategy [16]. Recently, we investigated the effects of the two contrary strategies on both external and internal loads compared to a control condition (free play) in female players. The findings show that the passive strategy leads to more distances covered at high acceleration and deceleration, and also to a higher heart rate, blood lactate concentration, and rating of perceived effort (1.1- to 7.2-fold of the smallest worthwhile change). Additionally, the active strategy leads to a lower blood lactate concentration (-2.4-fold), but once more to a higher rating of perceived effort (2.4-fold) [4]. Overall, the previous studies show that passive and active playing strategies have an impact on external and internal loads in tennis. Unfortunately, both previous studies [4, 16] failed to examine the effect of mixed playing strategy conditions in which passive competed against active players. Further, the vertical work performed during split steps, jumps, or services was not considered to quantify external loads [7]. Additionally, capillary blood lactate and portable respiratory gas procedures were used to investigate metabolic loads. However, during intermittent sports like tennis, blood lactate measures do not reflect metabolic situations at muscular levels [24], are invasive [25], less reliable [26], and allow no real-time monitoring [21]. Concerning the use of portable gas analyzers, it is self-explanatory that they interfere with maximal performances [27]. Last, due to the direct impact of playing strategies on technical-tactical actions [15], it is promising to also investigate the latter during the rallies, which has not been conducted so far. Thus, a study to address all these points is required; especially, in the hitherto barely studied females. This study aimed to investigate the effects of passive, active, and mixed playing strategies on external and internal loads in female tennis players during match play. Also, the underlying effects on the technical-tactical actions and activity profiles were examined. Based on previous studies [4, 16], it was hypothesized that each playing strategy has a different impact on the external and internal loads, technical-tactical actions, and activity profiles of the players during play. Here, we re-analyze few data of an already published study [4] and evaluate them together with comprehensive unpublished data. Compared to the published study, we (i) analyze not only passive and active, but also mixed playing strategy conditions, (ii) use inertial measurement units to determine external loads more complete, (iii) apply a new metabolic approach, and (iv) quantify ball placements during the rallies as an indicator of the technical-tactical behavior. Our outcomes may increase the understanding of playing strategies in female tennis players, which may be helpful for practical purposes like match analyses and training procedures. Materials and methods Participants and ethics statement To test our hypothesis, 12 well-trained female tennis players from local clubs were recruited. The inclusion and exclusion criteria were (i) an age of 20–30 years, (ii) a regional ranking ≤10, (iii) a right-handed stroking technique, and (iv) a balanced combination of baseline play and attacking toward the net. The players were excluded, if there was (v) an acute disease speaking against maximal load testing or (vi) no signed consent to participate. The players were informed of the purposes, procedures, and potential risks of the study. All procedures were pre-approved by the Ethics Committee of the University of Wuppertal (MS/JE 29.11.11) and were conducted in accordance with the Declaration of Helsinki. Table 1 summarizes the anthropometric characteristics and tennis backgrounds of the players. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. Anthropometric characteristics and tennis backgrounds of the female players (n = 12). https://doi.org/10.1371/journal.pone.0239463.t001 Experimental design All procedures were conducted on two sessions within one week during the last month of the outdoor season. The players were asked to report to both sessions well rested, to refrain from strenuous exercise for the prior 24 hours, and to prepare themselves as they would for an official competition. On the first session, the players were examined on a motorized treadmill for maximal oxygen uptake and heart rate in the laboratory. On the second session, the players were tested on court. The data collection took place outdoors on a red-clay court at 22–26°C and 38–45% humidity. After the players had warmed-up for 10 min with ground strokes, volleys, overhead strokes, and serves, they were asked to play points against an opponent of similar ability. The opponents were matched by a professional tennis coach, who was familiar with the players. The service changed after one player had served from both sides. During play, the players retrieved their own balls and counted the won points. The players played points over five playing conditions per 10 min. The conditions were separated by 5 min rest periods. Before each condition, the players were instructed to apply either: (i) a passive, (ii) an active, or (iii) their own playing strategy to succeed, as previously conducted [4]. The instructions were given on cards and formulated in an open manner. They read: (i) “try to win the points through a reduction of your own errors”, (ii) “try to win the points directly by yourself or by forcing the opponent to errors”, and (iii) “try to win the points with your own strategy that you would also apply in a competition”. The players were instructed not to communicate about the content of their received instructions. While the players applied the same strategies during three conditions (labeled as “both own”, “both passive”, and “both active”), they applied mixed strategies during two conditions (labeled as “mixed passive” and “mixed active”). For the mixed conditions, the data were analyzed from the perspectives of the passive and active players, respectively. Each pair of players played the five conditions in a randomized order, whereas the condition with the own strategy was always played first and served as control. After 5 min of play, the players were verbally reminded between two rallies by the tennis coach to consider their attained instructions. The last rally was played out. To estimate the external and internal loads, and also the underlying technical-tactical actions and activity profiles, global positioning system, inertial measurement unit, short-range telemetry, capillary blood, visual analog scale, and video camera procedures were applied. While the capillary blood and visual analog scale procedures were applied at the beginning of the rest periods, all other procedures were continuously applied during the five playing conditions. Fig 1 displays the design of the on-court testing session. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Design of the on-court testing session. https://doi.org/10.1371/journal.pone.0239463.g001 An overview of the data that were operationalized to reflect the external and internal loads, technical-tactical actions, and activity profile of the female tennis players during match play is also given. Laboratory testing To determine the maximal oxygen uptake and heart rate, an incremental running test on a motorized treadmill (H/P Cosmos, Pulsar, Nussdorf-Traunstein, Germany) was performed, as previously described [28]. Briefly, during the test, respiratory gas exchanges and heart rates were measured using an open-circuit breath-by-breath gas analyzer (Ganshorn, PowerCube-Ergo, Niederlauer, Germany) and short-range telemetry (Polar, T31, Kempele, Finland), respectively. The collected data were averaged over 10 s. Before each test, the gas analyzer was calibrated with a calibration gas (15.5% O2, 5% CO2 in N; Messner, Switzerland) and a precision 1-liter syringe (Ganshorn, Germany) according to the instructions of the manufacturer. The initial two increments of the protocol consisted of running for 4 min at 10 km/h at an inclination of 1% and 5%, respectively. Thereafter, the speed was increased every 2 min by 1 km/h until exhaustion was reached. The exhaustion was considered to be reached, if a plateau in oxygen uptake (increase <2 ml/kg/min) despite an increase in the workload was observed. Otherwise, three of the following four criteria had to be fulfilled: (i) a heart rate ≥95% of an age-predicted maximal heart rate (220-age), (ii) a respiratory exchange ratio ≥1.15, (iii) a capillary blood lactate concentration ≥8.0 mmol/l, or (iv) a Borg rating of perceived exertion ≥19. The maximal oxygen uptake and heart rate were defined as the highest recorded data during the workload. Before testing, anthropometric and body composition data were assessed. The body fat was predicted by a 4-point bioelectric impedance analysis (Bodystat, QuadScan 4000, Douglas, United Kindgom) in supine position. On-court testing External loads. To estimate the external loads, portable combined 10 Hz global positioning systems and 100 Hz inertial measurement units (MinimaxX S4, Catapult Innovations, Melbourne, Australia) were used. The global positioning system measured running speed data by the Doppler-shift, whereas the inertial measurement unit assessed acceleration data in all three movement planes by triaxial accelerometers [29]. The devices were worn beneath the players’ attire in custom-made neoprene harnesses located between the scapulae. To allow for the satellite lock, the devices were activated 15 min prior to the data collection. The collection took place under a cloudless sky between 09.00–11.00 A.M. and there were no tall buildings or trees around the court that could have had negative influences on signal qualities. During play, the devices had connections with ≥9 satellites and the horizontal dilution of position was ≤0.92, indicating ideal measurement conditions [30]. The reported raw data were exported from the proprietary software (Sprint, version 5.1.4, Catapult Innovations, Melbourne, Australia) to custom-made spreadsheets that incorporated macro-based calculations (Microsoft, Excel 2016, Redmond, WA, USA) for further analyses. To minimize noise, the running speed data assessed by the global positioning systems were proceeded by a second order low-pass Butterworth filter using a cut-off frequency of 1 Hz and two passes. From the filtered speed data and their integration over the time, the total running distances and those with low (<3 m/s) and high speed (≥3 m/s) were computed. Additionally, the filtered speed data were deviated over the time to derive acceleration and deceleration data. To also minimize noise from these data, they were filtered by a slightly modified filter (instead of 1 Hz with 0.5 Hz). Then, from the filtered data, the running distances with low (<2 m/s²) and high acceleration (≥2 m/s²) as well as low (>-2 m/s²) and high deceleration (≤-2 m/s²) were computed. Finally, from the acceleration data measured by the inertial measurement units in all three movement planes, the accumulated player loads were computed. The player load is one of the most common accelerometer derived parameter used to quantify total acceleration based external loads [31] and was calculated according to the following equation (Eq 1) [29]: (1) , whereas fwd, side, and up is forward, sideway, and upward acceleration, respectively. All filtering techniques, threshold definitions, and computational steps were applied according to previous tennis research [6]. Internal loads. To estimate the internal loads, heart rate, capillary blood lactate concentration, rating of effort, and metabolic power data were examined. The heart rate data were assessed at 1 Hz by short-range telemetry (Polar, T31, Kempele, Finland). The collected data were analyzed in relation to the maximal values achieved during the laboratory treadmill tests. They were considered as low (<85%) and high heart rates (≥85%) according to previous tennis research [32]. The lactate concentrations were determined from 20 μl capillary blood samples by an electro-enzymatic analyzer (EKF-diagnostics, Biosen C_line Sport, London, United Kingdom). From each blood sample, the concentrations were determined in duplicate and the mean was recorded. The ratings of efforts were quantified by 100 mm visual analogue scales. Beside these standard procedures to estimate internal loads in tennis, metabolic power data were also investigated. The metabolic power is defined as the instantaneous muscular energy demand that is required to maintain the ATP level constant [33]. To compute, the filtered running speed, acceleration, and deceleration data collected by the global positioning systems were used for an equivalent slope approach, as described in detail elsewhere [34]. Briefly, accelerated and decelerated running on a horizontal level is energetically equivalent to uphill and downhill running at a constant speed on an equivalent slope, whereby the slope is dictated by the forward acceleration and deceleration. Since the energetics of uphill and downhill running are well known and energy costs are independent of the speed and cluster on average about 4.0 J/kg/m, the energy costs of accelerated and decelerated running on a horizontal level can be estimated. The energy cost can then be multiplied by the underlying speed to calculate the instantaneous metabolic power [34]. For all computational steps, the original equations were used [35] that can be simplified as follows (Eq 2): (2) , whereas ES is equivalent slope, 4.0 is energy cost for running at a constant speed in J/kg/m, a is forward acceleration, g is acceleration due to gravity, KT is a terrain constant of 1.29, and v is speed. The equivalent slope was computed accordingly (Eq 3) [34]: (3) , whereas tan and arctan are tangent and arctangent, respectively. The computed metabolic power data were applied to the subsequent analyses: First, from the time integral of the metabolic power data, the energy expenditures were computed. Then, the energy expenditures spent with low and high metabolic power were calculated [36]. In this context, we [7] and a further research group [37] previously analyzed metabolic power data during tennis match play based on absolute thresholds. This can be seen as a limitation, because it is clear that metabolic capacities differ between the players. Thus, there is a need to develop individualized metabolic power thresholds [33]. Since metabolic power data can be converted into oxygen uptake units, one rational possibility for an individualization is to analyze the metabolic power data with respect to the maximal power of the aerobic system–that is the maximal oxygen uptake [36]. To implement this approach for the first time, the maximal oxygen uptakes achieved during the laboratory treadmill tests were converted into corresponding individual metabolic power thresholds as follows (Eq 4): (4) , whereas individual metabolic power threshold is in W/kg and maximal oxygen uptake is in ml/kg/min. The conversion assumes a respiratory exchange ratio of 0.96 for intermittent sports like tennis and a corresponding energy equivalent of 20.9 kJ per liter of oxygen uptake [36]. Plausibly, the outcomes of our individualization allow to differ between energy supplies from predominantly aerobic (metabolic power -2 m/s²) and high deceleration (≤-2 m/s²) were computed. Finally, from the acceleration data measured by the inertial measurement units in all three movement planes, the accumulated player loads were computed. The player load is one of the most common accelerometer derived parameter used to quantify total acceleration based external loads [31] and was calculated according to the following equation (Eq 1) [29]: (1) , whereas fwd, side, and up is forward, sideway, and upward acceleration, respectively. All filtering techniques, threshold definitions, and computational steps were applied according to previous tennis research [6]. Internal loads. To estimate the internal loads, heart rate, capillary blood lactate concentration, rating of effort, and metabolic power data were examined. The heart rate data were assessed at 1 Hz by short-range telemetry (Polar, T31, Kempele, Finland). The collected data were analyzed in relation to the maximal values achieved during the laboratory treadmill tests. They were considered as low (<85%) and high heart rates (≥85%) according to previous tennis research [32]. The lactate concentrations were determined from 20 μl capillary blood samples by an electro-enzymatic analyzer (EKF-diagnostics, Biosen C_line Sport, London, United Kingdom). From each blood sample, the concentrations were determined in duplicate and the mean was recorded. The ratings of efforts were quantified by 100 mm visual analogue scales. Beside these standard procedures to estimate internal loads in tennis, metabolic power data were also investigated. The metabolic power is defined as the instantaneous muscular energy demand that is required to maintain the ATP level constant [33]. To compute, the filtered running speed, acceleration, and deceleration data collected by the global positioning systems were used for an equivalent slope approach, as described in detail elsewhere [34]. Briefly, accelerated and decelerated running on a horizontal level is energetically equivalent to uphill and downhill running at a constant speed on an equivalent slope, whereby the slope is dictated by the forward acceleration and deceleration. Since the energetics of uphill and downhill running are well known and energy costs are independent of the speed and cluster on average about 4.0 J/kg/m, the energy costs of accelerated and decelerated running on a horizontal level can be estimated. The energy cost can then be multiplied by the underlying speed to calculate the instantaneous metabolic power [34]. For all computational steps, the original equations were used [35] that can be simplified as follows (Eq 2): (2) , whereas ES is equivalent slope, 4.0 is energy cost for running at a constant speed in J/kg/m, a is forward acceleration, g is acceleration due to gravity, KT is a terrain constant of 1.29, and v is speed. The equivalent slope was computed accordingly (Eq 3) [34]: (3) , whereas tan and arctan are tangent and arctangent, respectively. The computed metabolic power data were applied to the subsequent analyses: First, from the time integral of the metabolic power data, the energy expenditures were computed. Then, the energy expenditures spent with low and high metabolic power were calculated [36]. In this context, we [7] and a further research group [37] previously analyzed metabolic power data during tennis match play based on absolute thresholds. This can be seen as a limitation, because it is clear that metabolic capacities differ between the players. Thus, there is a need to develop individualized metabolic power thresholds [33]. Since metabolic power data can be converted into oxygen uptake units, one rational possibility for an individualization is to analyze the metabolic power data with respect to the maximal power of the aerobic system–that is the maximal oxygen uptake [36]. To implement this approach for the first time, the maximal oxygen uptakes achieved during the laboratory treadmill tests were converted into corresponding individual metabolic power thresholds as follows (Eq 4): (4) , whereas individual metabolic power threshold is in W/kg and maximal oxygen uptake is in ml/kg/min. The conversion assumes a respiratory exchange ratio of 0.96 for intermittent sports like tennis and a corresponding energy equivalent of 20.9 kJ per liter of oxygen uptake [36]. Plausibly, the outcomes of our individualization allow to differ between energy supplies from predominantly aerobic (metabolic power -2 m/s²) and high deceleration (≤-2 m/s²) were computed. Finally, from the acceleration data measured by the inertial measurement units in all three movement planes, the accumulated player loads were computed. The player load is one of the most common accelerometer derived parameter used to quantify total acceleration based external loads [31] and was calculated according to the following equation (Eq 1) [29]: (1) , whereas fwd, side, and up is forward, sideway, and upward acceleration, respectively. All filtering techniques, threshold definitions, and computational steps were applied according to previous tennis research [6]. Internal loads. To estimate the internal loads, heart rate, capillary blood lactate concentration, rating of effort, and metabolic power data were examined. The heart rate data were assessed at 1 Hz by short-range telemetry (Polar, T31, Kempele, Finland). The collected data were analyzed in relation to the maximal values achieved during the laboratory treadmill tests. They were considered as low (<85%) and high heart rates (≥85%) according to previous tennis research [32]. The lactate concentrations were determined from 20 μl capillary blood samples by an electro-enzymatic analyzer (EKF-diagnostics, Biosen C_line Sport, London, United Kingdom). From each blood sample, the concentrations were determined in duplicate and the mean was recorded. The ratings of efforts were quantified by 100 mm visual analogue scales. Beside these standard procedures to estimate internal loads in tennis, metabolic power data were also investigated. The metabolic power is defined as the instantaneous muscular energy demand that is required to maintain the ATP level constant [33]. To compute, the filtered running speed, acceleration, and deceleration data collected by the global positioning systems were used for an equivalent slope approach, as described in detail elsewhere [34]. Briefly, accelerated and decelerated running on a horizontal level is energetically equivalent to uphill and downhill running at a constant speed on an equivalent slope, whereby the slope is dictated by the forward acceleration and deceleration. Since the energetics of uphill and downhill running are well known and energy costs are independent of the speed and cluster on average about 4.0 J/kg/m, the energy costs of accelerated and decelerated running on a horizontal level can be estimated. The energy cost can then be multiplied by the underlying speed to calculate the instantaneous metabolic power [34]. For all computational steps, the original equations were used [35] that can be simplified as follows (Eq 2): (2) , whereas ES is equivalent slope, 4.0 is energy cost for running at a constant speed in J/kg/m, a is forward acceleration, g is acceleration due to gravity, KT is a terrain constant of 1.29, and v is speed. The equivalent slope was computed accordingly (Eq 3) [34]: (3) , whereas tan and arctan are tangent and arctangent, respectively. The computed metabolic power data were applied to the subsequent analyses: First, from the time integral of the metabolic power data, the energy expenditures were computed. Then, the energy expenditures spent with low and high metabolic power were calculated [36]. In this context, we [7] and a further research group [37] previously analyzed metabolic power data during tennis match play based on absolute thresholds. This can be seen as a limitation, because it is clear that metabolic capacities differ between the players. Thus, there is a need to develop individualized metabolic power thresholds [33]. Since metabolic power data can be converted into oxygen uptake units, one rational possibility for an individualization is to analyze the metabolic power data with respect to the maximal power of the aerobic system–that is the maximal oxygen uptake [36]. To implement this approach for the first time, the maximal oxygen uptakes achieved during the laboratory treadmill tests were converted into corresponding individual metabolic power thresholds as follows (Eq 4): (4) , whereas individual metabolic power threshold is in W/kg and maximal oxygen uptake is in ml/kg/min. The conversion assumes a respiratory exchange ratio of 0.96 for intermittent sports like tennis and a corresponding energy equivalent of 20.9 kJ per liter of oxygen uptake [36]. Plausibly, the outcomes of our individualization allow to differ between energy supplies from predominantly aerobic (metabolic power