Effects of post-activation performance enhancement on tennis serve velocity and accuracy Authors: Wenlai Cui, Yujian Deng, Wenbo Zhang, Bo Pang Published: 2026 ABSTRACT AbstractThis study systematically compared the effects of different activation sites and modes on serve velocity (SV) and serve accuracy (SA) across multiple time windows to clarify the time course and practical significance of post-activation performance enhancement (PAPE). A 3 (activation site: upper limb, lower limb, full body) × 2 (activation mode: ballistic training, BT; high-resistance training, HRT) × 6 (time: PRE, 3, 6, 9, 12, 15 min) design was employed. SV was analyzed using repeated-measures ANOVA, and SA, which did not meet normality assumptions, was analyzed with generalized estimating equations (GEE). Percentage changes (Δ%) were calculated for confirmatory analysis. SV showed significant main effects of time (F = 212.33, p < 0.001, η2p = 0.910), activation mode (F = 34.57, p < 0.001, η2p = 0.622), and a Time × Activation mode interaction (F = 7.423, p < 0.001, η2p = 0.261), with BT producing higher SV overall. SVΔ% analysis indicated BT advantage in the early-to-mid window (3–9 min), with upper-limb and full-body BT exceeding lower-limb HRT (p < 0.01, Cohen’s d = 2.08–3.77). SA showed a significant time effect (Wald χ2 = 222.40, p < 0.001), but no differences were observed between activation conditions at the same time points (all p > 0.05). SV increased after activation, with BT producing greater improvements within 3–9 min. Upper-limb and full-body activation tended to outperform lower-limb activation. SA changed over time, without consistent between-condition differences.Keywords: PAPE, Activation site, Activation mode, Serve velocity, Serve accuracySubject terms: Health care, Medical research, Physiology Introduction Post-activation performance enhancement (PAPE) refers to the transient improvement in voluntary physical performance observed following high-intensity conditioning activities1–3. Although PAPE is often discussed alongside post-activation potentiation (PAP), the two concepts differ fundamentally in their underlying nature4. PAP is typically defined as a physiological phenomenon occurring at the muscular level, characterized by an acute enhancement of twitch force. Its classical mechanism is associated with the phosphorylation of myosin regulatory light chains, which increases the sensitivity of the contractile apparatus to Ca2⁺, thereby enabling greater force production at a given Ca2⁺ concentration, rather than directly increasing Ca2⁺ release5–8. In contrast, PAPE primarily reflects improvements in actual performance outcomes and is considered a multifactorial phenomenon involving both physiological and neural mechanisms. In addition to PAP-related processes, potential contributors include elevated muscle temperature, changes in muscle water content, and increased central nervous system excitability4,9. The interaction of these factors may enhance neuromuscular efficiency and motor unit recruitment, ultimately improving subsequent explosive performance. Previous research has demonstrated the effectiveness of PAPE in various explosive tasks, such as vertical jumping, sprinting, and throwing, typically manifesting as short-term enhancements in power output, speed, and force production10,11. In tennis, although some studies have begun to explore its effects on serve performance12–14, the available evidence remains limited, and the effectiveness of different activation strategies is still inconclusive, warranting further investigation. Tennis serve is the only offensive technique in tennis that is fully under the athlete’s voluntary control. It is also a highly explosive, multi-joint coordinated movement that integrates strength, speed, coordination, and precision. Its technical quality directly determines the probability of winning service games and gaining tactical advantage in matches14,15. Previous biomechanical studies using three-dimensional kinematic analysis have shown that serve velocity is significantly associated with lower-limb driving force, trunk rotational velocity, and upper-limb joint angular velocity. Among these factors, trunk rotational angular velocity exhibits a strong positive correlation with ball speed, while the temporal coordination among joints plays a decisive role in the efficiency of kinetic energy transfer16. From a dynamic perspective, the tennis serve demonstrates a typical “kinetic chain” transmission pattern. Related studies indicate that during high-level serving, the ground reaction force generated by lower-limb extension is sequentially transferred through hip and trunk rotation to the upper limbs. The momentum at each segment follows a characteristic pattern of “decreasing magnitude and increasing velocity” from proximal to distal segments17,18. Further research has shown that this energy transfer process has a clear temporal structure: the peak of energy output and the forward displacement of the center of mass typically occur in the pre-impact phase, enabling maximal velocity generation at ball contact19. Ultimately, the upper limb musculature completes the stroke through a rapid eccentric-to-concentric contraction transition, thereby effectively enhancing ball speed and hitting efficiency. Given the high demand of tennis serving on neuromuscular coordination and rapid force production capacity, previous studies have shown that changes in muscle activation levels can significantly affect motor unit recruitment efficiency and force production timing, thereby influencing serve speed and consistency20,21. PAPE has been demonstrated to transiently increase neuromuscular excitability and improve power output and movement speed in explosive tasks (e.g., jump height, sprint performance), although its effects vary depending on activation protocols and time windows22,23. Although the modulatory effects of PAPE on explosive motor performance provide a theoretical basis for optimizing tennis serve technique, existing research in this area remains limited and warrants further investigation. From a research scope perspective, previous studies on the application of PAPE in tennis have predominantly focused on the effects of pre-activation of specific body regions on general performance outcomes, such as movement speed and change-of-direction ability24,25. In contrast, research directly examining the PAPE effects on tennis serve performance is scarce and findings remain inconsistent. Therefore, the present study systematically compared the effects of different combinations of activation sites and activation modalities across multiple time windows on serve velocity (SV) and serve accuracy (SA), with the aim of clarifying the temporal characteristics of PAPE and providing theoretical support and practical guidance for the development of long-term training programs and pre-match activation strategies. Based on existing findings on PAPE and the kinetic chain characteristics of the tennis serve, the following hypotheses were proposed in this study: (1) different activation modalities have distinct effects on serve performance, with ballistic training (BT) being more effective than high-resistance training (HRT) in acutely improving SV; (2) different activation sites produce different levels of performance enhancement, with upper-limb and full-body activation being more beneficial than lower-limb activation for improving serve performance; (3) the PAPE effect on serve velocity is time-dependent and may be more pronounced during the early post-activation phase (e.g., 3–9 min); and (4) compared with serve velocity, SA shows a lower degree of responsiveness to different activation conditions. Methods Participants Twenty-four male high-level tennis players were recruited for this study (age: 23.21 ± 3.02 years; height: 180.46 ± 5.17 cm; body mass: 74.63 ± 8.77 kg; body mass index: 22.90 ± 2.43 kg·m⁻2; training experience: 10.46 ± 3.06 years). Inclusion criteria were as follows: (1) male tennis players aged between 18 and 30 years; (2) attainment of a top-three finish in provincial-level or higher competitions, ensuring adequate representativeness and competitive proficiency; (3) a minimum of two years of regular tennis training, with a training frequency of at least three sessions per week; and (4) voluntary participation with written informed consent provided prior to enrollment. Exclusion criteria included: (1) the presence of chronic diseases (e.g., cardiovascular disease, hypertension) or severe musculoskeletal disorders; and (2) a history of severe sports-related injury within the previous six months, particularly involving the upper-limb joints. The study protocol was approved by the Ethics Committee of the Capital University of Physical Education and Sports (approval no. 2024A158) and registered with the Chinese Clinical Trial Registry (ChiCTR2400090958). All procedures were conducted in accordance with the ethical standards of the World Medical Association (Declaration of Helsinki). Experimental design Athletes were randomly assigned to three groups based on different pre-activation sites: the upper-limb pre-activation group, lower-limb pre-activation group, and full-body pre-activation group (between-subject factor). On this basis, a within-subject crossover design was adopted, in which all participants completed two types of activation protocols in different experimental phases: BT and HRT (Fig. 1a). The order of the two activation modalities was randomized within subjects, with a one-week washout period between sessions to minimize potential carryover effects from the previous stimulus. Experimental testing procedure and timeline. The specific training protocols were as follows: upper-limb BT was performed using a seated forward medicine ball throw, whereas upper-limb HRT involved a front barbell press; lower-limb BT was performed as a weighted half squat jump, and lower-limb HRT consisted of a barbell half squat; full-body BT was performed using a standing forward medicine ball throw, while full-body HRT involved a barbell high pull. Detailed pre-activation protocols are presented in Table 1. Table 1Pre-activation content.Activation siteActivation modeActivation methodsLoad intensityWeight (kg)RepetitionsSetsUpper-limbBTSeated forward medicine ball throw30% 1RM25.63 + 4.4163HRTFront barbell press80% 1RM68.33 + 11.7633Lower -limbBTWeighted half squat jump30% 1RM33.60 + 6.9663HRTBarbell half squat80% 1RM89.60 + 18.5533Full -bodyBTStanding forward medicine ball throw30% 1RM22.08 + 5.3463HRTBarbell high pull80% 1RM58.87 + 14.2333 Pre-activation content. For movements such as medicine ball throws, which are not suitable for traditional one-repetition maximum (1RM) testing, an equivalent load estimation method was used to determine an intensity of approximately 30% 1RM. Based on each participant’s maximal strength level of the corresponding muscle groups (e.g., 1RM bench press for upper limbs or 1RM half squat for lower limbs), medicine ball mass was individually adjusted. In addition, to ensure consistency of training stimuli across conditions, a fixed rest interval of 2 min was provided between sets to minimize the potential influence of accumulated fatigue on subsequent performance. Testing procedures and outcome measures All participants completed one instructional practice session and 1RM strength test one week prior to the experiment to ensure proper execution of movement techniques during the formal testing and to determine the 30% and 80% 1RM loading intensities. The formal experimental procedures are shown in Fig. 1b. Measurements were conducted at pre-activation and at 3, 6, 9, 12, and 15 min post-activation, resulting in six total time points. At each time point, participants performed flat tennis serve tests, with SV and SA recorded as outcome measures. SV was measured using a Stalker PRO II handheld radar gun positioned 3 m behind the server at a height of 2.5 m, pointing in the direction of serve. The Stalker PRO II radar gun operates based on the Doppler principle, with a measurement range of 0–1448 km/h and an accuracy of approximately 3%. At each time point, participants performed a total of six flat tennis serves, including three serves to the deuce court and three serves to the ad court. During SV testing, only successful serves were recorded, and the best performance was used for subsequent analysis. The use of the best score rather than the mean was adopted based on common practices in assessing explosive performance. Previous studies have demonstrated that, in maximal power and speed tasks, the best performance more accurately reflects an individual’s true maximal capability and reduces potential underestimation caused by technical variability or occasional execution errors. In addition, this approach has been widely applied in studies on tennis serve velocity and similar explosive performance tasks (Fig. 2). Schematic illustration of the serve test. For SA testing, the same procedures and target area classification were applied to both service directions (deuce court and advantage court) to ensure consistency and comparability. Specifically, each half of the service court was divided into four equal zones. The area near the “T” was defined as the high-value target zone (2 points), while the remaining valid service areas were assigned 1 point, and out-of-bounds serves were scored as 0. At each time point, participants performed six serves in total (three to the deuce court and three to the ad court), with a maximum possible total score of 12 points. The cumulative SA score across the six serves was recorded and used for analysis26. Throughout the testing period, participants maintained their habitual lifestyle and regular diet, abstained from nutritional supplements and anti-inflammatory medications, and avoided caffeine intake for 24 h prior to each testing session27, as well as strenuous physical activity28. All tests were conducted at the same time of day to minimize potential effects of circadian rhythm variations on the study outcomes29. Statistical analysis An a priori power analysis was conducted using G*Power 3.1.9.5 (Heinrich Heine University Düsseldorf, Germany) to determine the required sample size for detecting the between–within interaction effect in a 3 × 2 × 6 mixed repeated-measures ANOVA. The following parameters were specified: effect size f = 0.25 (moderate), α = 0.05, statistical power (1 − β) = 0.90, number of between-subject levels (pre-activation groups) = 3, number of within-subject conditions (2 activation modalities × 6 time points) = 12, assumed correlation among repeated measures = 0.60, and nonsphericity correction ε = 1. The effect size of f = 0.25 was based on Cohen’s classification of a medium effect size30, and was further informed by previously reported medium-to-large effect sizes in studies on PAPE. This setting was adopted to ensure sufficient statistical power to detect practically meaningful effects within a reasonable sample size. The analysis indicated a required total sample size of 18 participants (6 per group). To account for potential data loss, an additional 2–3 participants per group were recruited. All statistical analyses were performed using SPSS version 26.0 (IBM, USA). Data are presented as mean ± standard deviation (Mean ± SD). Prior to analysis, the Shapiro–Wilk test was applied to assess data normality, and measurement reliability was evaluated using the intraclass correlation coefficient (ICC). For SV, a mixed-design repeated-measures analysis of variance (ANOVA) was conducted to examine the effects of activation site, activation mode, and time. Time (PRE, 3, 6, 9, 12, and 15 min post-activation) and activation mode (BT, HRT) were treated as within-subject factors, whereas activation site (upper limbs, lower limbs, full body) was treated as a between-subject factor. Mauchly’s test was used to assess the assumption of sphericity for within-subject factors, and Greenhouse–Geisser corrections were applied when this assumption was violated. To minimize the potential influence of baseline differences, activation mode and activation site were combined into six activation conditions, and percentage changes in SV (Δ%) were calculated for supplementary analyses using the formula: Δ% = (Tx − PRE) / PRE × 100%. When significant main effects or interactions were detected, Bonferroni-adjusted post hoc comparisons were performed, and effect sizes were calculated using Cohen’s d. Cohen’s d was calculated based on post hoc pairwise comparisons by dividing the mean difference between two conditions by the pooled standard deviation, in order to reflect the practical significance of between-condition differences. Effect size magnitudes were interpreted according to the criteria proposed by Hopkins et al.31: small (> 0.2 and < 0.6), moderate (≥ 0.6 and < 1.2), large (≥ 1.2 and < 2.0), and very large (≥ 2.0). Because SA data did not meet normality assumptions, generalized estimating equations (GEE) were used to examine changes in SA over time under different activation conditions. In the GEE models, participants were specified as the repeated-measures unit, with time included as a within-subject variable and activation mode, activation site, and their interaction entered as fixed effects. The model assumed that SA followed a Gaussian distribution and employed an identity link function. An autoregressive correlation structure of order 1 (AR(1)) was specified to account for within-subject correlations across time points. Model parameters were estimated using robust standard errors, and Wald χ2 statistics were used to evaluate the significance of main and interaction effects. When significant effects were observed, post hoc pairwise comparisons were conducted based on estimated marginal means, with Bonferroni adjustment applied to control for multiple comparisons. For percentage changes in SA (Δ%) that satisfied normality assumptions, a two-factor repeated-measures ANOVA was further performed as a confirmatory analysis, with Bonferroni correction applied for significant effects. Results Serve velocity All variables were normally distributed and showed excellent within-session reliability (ICC_single = 0.902, ICC_average = 0.991); Greenhouse–Geisser corrections were applied to within-subject effects due to violations of sphericity for Time and Time × activation mode. The repeated-measures ANOVA revealed a significant main effect of time on SV (F = 212.326, p < 0.001, η2p = 0.910). A significant main effect of activation mode was also observed (F = 34.574, p < 0.001, η2p = 0.622), with SV being significantly higher under BT than under HRT. The main effect of activation site was not significant (F = 0.900, p = 0.422). Regarding interaction effects, a significant Activation mode × Time interaction was detected (F = 7.423, p < 0.001, η2p = 0.261). In addition, the Time × Activation site interaction reached statistical significance (F = 6.421, p < 0.001, η2p = 0.379); however, post hoc comparisons did not reveal consistent significant differences between activation sites at individual time points. Neither the Activation mode × Activation site interaction nor the three-way interaction reached statistical significance (all p > 0.05). These findings indicate that the effects of different activation sites on serve velocity are primarily reflected in temporal pattern differences over time, rather than stable between-group differences at specific time points. In other words, different activation sites may vary in the onset timing and magnitude of the PAPE response; however, such differences were not sufficient to produce significant group effects at any single time point. This pattern is commonly observed in repeated-measures studies involving multiple time points, particularly when effects evolve gradually over time and inter-individual variability is relatively large. Post hoc analyses indicated that SV was significantly higher at 6, 9, 12, and 15 min post-activation compared with PRE and 3 min post (all p < 0.05), whereas no significant difference was observed between 12 and 15 min post. Further within-condition effect size analyses showed that, relative to PRE, SV under BT was significantly increased at 12 and 15 min post, with moderate effect sizes (Cohen’s d = 0.67 and 0.65, respectively). In contrast, effect sizes under HRT were generally smaller, reaching only small to moderate magnitudes at 12 and 15 min post (Cohen’s d = 0.57–0.63). At identical time points, no stable significant differences were observed among different combinations of activation site and activation mode. These findings suggest that, although a significant main effect of activation mode was observed, this effect was primarily driven by overall differences across time. Specifically, the BT condition exhibited generally higher serve velocity across multiple time points, rather than a consistent and significant advantage at any single time point. In other words, the superiority of BT over HRT appears to be cumulative or global in nature; however, at individual time points, the magnitude of the difference was insufficient to reach statistical significance. SV data are shown in Table 2. Table 2Serve velocity (mean ± SD) across time and activation conditions.Activation conditionsPRE3 min6 min9 min12 min15 minUpper-limb BT162.26 ± 14.13163.79 ± 12.83166.89 ± 12.98171.08 ± 13.76170.99 ± 12.42170.44 ± 12.74Upper-limb HRT162.36 ± 13.79160.64 ± 12.31164.21 ± 12.44167.93 ± 12.82169.35 ± 12.55169.95 ± 13.04Lower-limb BT158.64 ± 13.82159.05 ± 13.85159.95 ± 13.43162.50 ± 14.76165.21 ± 13.54165.29 ± 13.09Lower-limb HRT158.54 ± 13.76158.81 ± 13.67158.56 ± 13.94160.44 ± 13.94164.25 ± 12.89165.20 ± 12.91Full-body BT164.66 ± 11.79165.98 ± 11.39170.34 ± 9.70173.96 ± 11.27175.28 ± 11.46174.48 ± 11.45Full-body HRT164.81 ± 11.75163.25 ± 11.80166.85 ± 12.25169.23 ± 11.13173.04 ± 9.07174.09 ± 9.91 Serve velocity (mean ± SD) across time and activation conditions. The repeated-measures ANOVA results for SVΔ% indicated a significant main effect of time (F = 157.119, p < 0.001, η2p = 0.789). In addition, a significant Time × activation condition interaction effect was observed (F = 3.375, p < 0.001, η2p = 0.287). Bonferroni-adjusted post hoc comparisons indicated that differences among activation conditions were primarily concentrated within the early-to-mid time window (3–9 min post-activation), with an overall advantage observed under BT compared with HRT. At 3 min post-activation, SV under the upper-limb BT and full-body BT conditions was significantly higher than that under the upper-limb HRT and full-body HRT conditions (all p ≤ 0.030), with large to very large effect sizes (Cohen’s d = 1.58–2.43). At 6 min post-activation, upper-limb BT and full-body BT continued to elicit significantly higher SV compared with the lower-limb HRT condition (all p ≤ 0.031), with extremely large effect sizes (Cohen’s d = 2.67–3.17). At 9 min post-activation, the between-condition differences were most pronounced, with SV under the upper-limb BT and full-body BT conditions being significantly higher than that under the lower-limb HRT and full-body HRT conditions (all p ≤ 0.049; Cohen’s d = 2.08–3.77). It should be noted that some of the Cohen’s d effect sizes were relatively large, which is primarily attributable to the use of Δ% as the outcome measure in this study. Δ% may amplify between-condition differences when baseline values are low or inter-individual variability is limited, thereby inflating standardized effect sizes. Therefore, these results should be interpreted in conjunction with analyses of raw values and η2p to provide a more comprehensive understanding. In contrast, no statistically significant differences were detected among activation conditions at 12–15 min post-activation (p > 0.05); therefore, effect sizes were not further reported for these time points. Time-course changes in SV and SVΔ% are shown in Fig. 3. Time-course changes in serve velocity (SV) and SVΔ% under different activation conditions. Serve accuracy GEE were used to examine changes in SA over time under different activation conditions. The results revealed a significant main effect of time (Wald χ2 = 222.400, df = 5, p < 0.001) and a significant main effect of activation site (Wald χ2 = 10.284, df = 2, p = 0.006), whereas the main effect of activation mode was not significant (Wald χ2 = 1.692, df = 1, p = 0.193). The interaction analysis showed a significant interaction between activation mode and time (Wald χ2 = 32.898, df = 5, p < 0.001), indicating that the temporal pattern of changes in serve accuracy differed across activation modalities. In contrast, the interaction between activation site and time was not significant (Wald χ2 = 6.779, df = 10, p = 0.746), suggesting that different activation sites had limited influence on the temporal evolution of serve accuracy. The interaction between activation site and activation mode approached statistical significance (Wald χ2 = 4.678, df = 2, p = 0.096). However, the three-way interaction among activation site, activation mode, and time was not significant (Wald χ2 = 10.246, df = 10, p = 0.419). Pairwise comparisons indicated that significant differences were primarily attributable to within-condition changes across time, whereas no consistent significant differences were observed among different combinations of activation site and activation mode at the same time points. SA data are shown in Table 3. Table 3Serve accuracy (mean ± SD) across time and activation conditions.Activation conditionsPRE3 min6 min9 min12 min15 minUpper-limb BT6.50 ± 0.937.88 ± 2.429.00 ± 1.858.25 ± 1.588.75 ± 1.288.25 ± 1.04Upper-limb HRT6.50 ± 0.767.25 ± 1.288.38 ± 1.068.75 ± 1.049.75 ± 1.499.13 ± 1.73Lower-limb BT5.88 ± 0.838.50 ± 0.768.38 ± 1.608.63 ± 0.748.38 ± 0.748.50 ± 0.93Lower-limb HRT6.13 ± 1.137.38 ± 1.307.75 ± 1.048.38 ± 1.069.25 ± 1.048.75 ± 1.67Full-body BT6.75 ± 1.399.13 ± 0.839.63 ± 1.309.38 ± 2.008.88 ± 1.468.38 ± 1.19Full-body HRT6.13 ± 0.996.88 ± 0.997.88 ± 1.558.63 ± 1.4110.38 ± 1.199.00 ± 1.07 Serve accuracy (mean ± SD) across time and activation conditions. Because the percentage change in SAΔ% satisfied the assumption of normality, a 6 (activation condition) × 5 (time) repeated-measures ANOVA was conducted. Mauchly’s test of sphericity indicated that the time factor met the sphericity assumption (W = 0.808, p = 0.472); therefore, results were reported using uncorrected degrees of freedom. The analysis revealed a significant main effect of time (F(4, 168) = 6.33, p < 0.001, partial η2 = 0.131), whereas the main effect of activation condition was not significant (F(5, 42) = 0.42, p = 0.832, partial η2 = 0.048). The Time × activation condition interaction reached statistical significance (F(20, 168) = 1.79, p = 0.025, partial η2 = 0.175). Post hoc analyses showed that, under HRT conditions, SA at 12 min was significantly higher than at 3 min post-activation. Specifically, SA under upper-limb HRT was significantly greater at 12 min than at 3 min post-activation (p = 0.007, Cohen’s d = 1.34); similarly, lower-limb HRT demonstrated a significant increase at 12 min compared with 3 min post-activation (p = 0.044, Cohen’s d = 0.97). In addition, the full-body HRT condition exhibited the most pronounced within-condition improvement, with SA at 12 min being significantly higher than at 3 min post-activation (p < 0.001, Cohen’s d = 1.68). However, at any individual time point, no statistically significant differences in SAΔ% were observed among activation conditions following Bonferroni correction (all p > 0.05). Time-course changes in SA and SAΔ% are shown in Fig. 4. Time-course changes in serve accuracy (SA) and SAΔ% under different activation conditions. Discussion A 3 × 2 × 6 experimental design was employed to systematically examine the effects of different combinations of activation sites and activation modalities on SV and SA. In addition, relative changes in SV and SA were calculated to compare differences among the six activation conditions. The present findings demonstrate that pre-activation interventions exert a pronounced time-dependent effect on SV, with clear differences in both the temporal profile and magnitude of the effects between activation modalities. Overall, SV did not peak immediately following activation but increased significantly within 6–15 min post-activation. This suggests that, even in a sport-specific movement such as the tennis serve—which is characterized by multi-joint coordination, kinetic chain transfer, and high demands on temporal coordination—the PAPE effect also requires an adequate recovery window to achieve an optimal balance between fatigue dissipation and enhanced neuromuscular excitation1–3. These results are consistent with previous observations of delayed PAPE effects in explosive tasks such as jumping and sprinting and further support the notion that PAPE is not an instantaneous phenomenon, but rather a physiologically regulated process with distinct temporal characteristics11,32. In terms of activation modality, the BT condition was overall significantly superior to HRT, with its advantages being primarily concentrated in the early to mid post-activation phase (3–9 min). Under the BT condition, large effect sizes were observed at multiple time points, indicating a statistically meaningful facilitation effect on serve velocity. However, given that the effect sizes were calculated based on Δ%, which may inflate between-condition differences when baseline variability is small, these results should be interpreted with caution and considered alongside raw data and overall effect sizes for a more comprehensive evaluation. From a mechanistic perspective, these findings may be explained by the explosive, high-velocity nature emphasized in BT. Compared with HRT, BT may more effectively enhance motor unit recruitment rate, discharge frequency, and the rapid responsiveness of the neuromuscular system, thereby more directly supporting the velocity-oriented demands of the tennis serve action33. In contrast, although HRT elicited moderate improvements in SV at later time points (12–15 min), its overall effect sizes were smaller, suggesting that the higher fatigue induced by heavy-load stimuli may constrain early performance, with potential PAPE benefits emerging only after more prolonged recovery34,35. Accordingly, differences between activation modalities diminished and became non-significant at 12 and 15 min, implying a gradual convergence of neuromuscular states over time. Although the Activation site × Time interaction reached statistical significance, no consistent and stable velocity advantage of any activation site was observed at individual time points. This finding suggests that the significant interaction more likely reflects differences in the temporal response patterns of serve velocity across activation sites (i.e., divergent time-course trajectories), rather than stable between-group differences at specific time points. Further analyses based on SVΔ% showed that during the early to mid phase of the PAPE effect (3–9 min), the upper-limb BT and full-body BT conditions repeatedly exhibited significantly higher serve velocities than the lower-limb HRT condition, with relatively large effect sizes. This indicates that, within a specific temporal window, upper-limb and full-body activation may be more effective in translating the PAPE effect into performance gains in serve velocity. It should be noted that these effect sizes were derived from Δ% values and may therefore overestimate between-condition differences to some extent; thus, their practical significance should be interpreted in conjunction with the raw data. From a mechanistic perspective, this phenomenon may be related to the proximal-to-distal acceleration characteristics of the tennis serve. Previous studies have shown that final ball velocity is primarily determined by the angular velocity of the upper-limb–racket system and the distal segment linear velocity, whereas the lower limbs mainly contribute to force transmission and postural stabilization, exerting only an indirect influence on terminal velocity36,37. Therefore, under acute PAPE conditions, upper-limb and full-body activation may more directly enhance serve velocity by improving neuromuscular coordination and distal segment control efficiency. However, as no direct biomechanical or neuromuscular measurements were included in the present study, these interpretations remain speculative and should be verified in future research using electromyography or kinematic analyses. Moreover, the observed advantage does not appear to be attributable to a single activation site in a stable manner, but rather reflects a systemic effect jointly modulated by activation modality and recovery time. In addition, the repeated serve assessments across multiple post-activation time points may have introduced potential familiarization, task adaptation, or learning effects, which could partially influence subsequent serve performance. Although participants completed a familiarization session prior to formal testing and standardized procedures were maintained throughout the experiment, these potential effects cannot be entirely excluded and should be considered when interpreting the findings. With respect to serve accuracy, the GEE analysis revealed significant main effects of Time and Activation site, as well as several interaction effects. However, post hoc pairwise comparisons indicated that these significant effects were primarily driven by temporal changes within the same activation condition, rather than by differences between activation conditions at identical time points. Consistently, further analysis of SAΔ% confirmed that, at any given time point, changes in accuracy did not differ significantly among activation conditions, indicating that the choice of activation site or activation modality does not confer a meaningful advantage for enhancing SA. Notably, under the HRT condition, moderate to large within-condition improvements in SA were observed at 12 min, suggesting that following sufficient recovery, higher-load stimuli may positively influence accuracy by increasing strength reserve capacity and stabilizing neuromuscular control38. The present findings show both consistencies and discrepancies with previous studies on PAPE in tennis serving. Consistent with the work of Gelen, our results support the conclusion that dynamic movements and upper-limb explosive actions facilitate improvements in SV39. Similarly, our findings align with those of Baiget et al., who reported that joint-specific maximal contractions enhanced SV without impairing accuracy14. Moreover, the present study further demonstrates that, under specific conditions, full body and upper-limb activation are more effective than lower-limb activation, likely due to the multi-joint and coordinative nature of the tennis serve. In contrast, Terraza-Rebollo et al. reported no effect of HRT on SV in adolescent tennis players, although SA remained unaffected40. While our findings regarding SA are consistent with their results, we observed significant PAPE-induced enhancements in SV within specific temporal windows. These discrepancies may be explained by differences in participant characteristics, experimental protocols, and performance assessment timing. Taken together, the present findings indicate that the performance-enhancing effects of PAPE on tennis serving are not determined by a single stimulus type or localized activation alone. Rather, they depend on the interaction between activation modality, movement specificity, and the timing of performance assessment. By employing a multi-time-point design, this study provides a more nuanced understanding of the temporal characteristics of PAPE under different activation strategies and offers a framework for reconciling inconsistencies in previous research. Although the present study provides valuable evidence regarding the effects of PAPE on tennis serve performance, several limitations should be acknowledged. First, only male high-level tennis players were included, resulting in a relatively homogeneous sample. Therefore, the generalizability of the findings to players of different performance levels, genders, or age groups remains to be further examined. Second, only the flat serve was investigated, without consideration of other serve types such as topspin or slice serves. Given the differences in kinetic chain structure and force production patterns across serve techniques, the generalizability of the present conclusions may be limited. In addition, although the results were interpreted from biomechanical and neuromuscular perspectives, no direct measurements (e.g., electromyography or kinematic data) were collected. Therefore, interpretations regarding the underlying mechanisms remain speculative and should be treated with caution.From a practical perspective, the findings suggest that explosive activation strategies (e.g., BT) may be beneficial for acutely enhancing serve velocity within a short temporal window (approximately 3–9 min) in pre-competition or training settings. However, practitioners should consider individual differences, recovery intervals, and specific competitive contexts when applying these strategies, and tailor both activation modalities and timing accordingly. Finally, as the present study did not account for complex match-related factors such as accumulated fatigue or psychological stress, its practical applicability should be further validated in more ecologically valid, competition-like settings. Conclusion Pre-activation interventions exhibited a clear time-dependent effect on tennis serve velocity, and this effect did not peak immediately after activation but gradually emerged and reached relatively high levels between 6 and 15 min post-activation. This finding suggests that, in the context of tennis serving—a complex skill involving multi-joint coordination and kinetic chain transmission—the PAPE effect requires an adequate recovery period to achieve a dynamic balance between fatigue dissipation and enhanced neuromuscular excitation. In terms of activation modality, the BT condition showed an overall superior trend compared with HRT, particularly within the 3–9 min post-activation window. However, this advantage was not consistently significant across all time points. Although relatively large effect sizes were observed at several time points, these results should be interpreted with caution, as they were derived from SVΔ% and should be considered alongside raw data and overall effect sizes for a more robust interpretation. Further analyses indicated that during the early to mid phase of the PAPE response, both upper-limb and full-body BT conditions demonstrated higher serve velocities than the lower-limb HRT condition at several time points. This suggests that, within a specific temporal window, different combinations of activation modality and activation site may influence the efficiency of translating pre-activation stimuli into performance gains in serve velocity. However, this advantage was not consistently observed across all time points, and therefore does not provide sufficient evidence to support a stable superiority of any specific activation site. Regarding SA, significant effects of time and activation site were observed; however, no significant differences were detected between activation conditions at any given time point. Complementary analyses of SAΔ% confirmed that activation condition did not meaningfully influence changes in accuracy across time, thereby supporting the notion that high-intensity pre-activation does not compromise technical stability in tennis serve performance. Acknowledgements We would like to thank the participants for their participation in this project. Author contributions Wenlai Cui was responsible for study conception, experimental design, data collection, and manuscript drafting. Bo Pang supervised the research framework and critically revised the manuscript. Yujian Deng and Wenbo Zhang contributed to data collection and statistical analyses. Funding This study was supported by the Fundamental Research Funds for the Central Universities (Project title: Multimodal Assessment and Targeted Exercise Intervention Mechanisms for the Physical Fitness Development of Young Police Officers in the New Era; Project No.: 2026JKF02SK10) and by the National Team Scientific Research Key Project (Project title: Research on a Championship Model for Elite Tennis Players’ Performance Empowered by Artificial Intelligence; Project No.: 24KJCX003). Data availability The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.