**3. Results**

#### *3.1. Search Results*

An initial search identified 226 studies. After pooling the results and removing duplicates, 136 articles were screened for titles and abstracts. Finally, there were 29 studies successfully meeting the eligibility criteria (Figure 1). The studies were excluded because they were irrelevant (n = 30); they involved players with disabilities, musculoskeletal problems, or children (n = 9); they used robotic players, simulations or theoretical calculations (n = 26); they ocused on psychological issues, tactics, decision-making, coaching, cardiopulmonary or metabolic assessments (n = 32); they were survey, conference paper, review, and expert comment papers (n = 6). One study did not fall into the inclusion criteria of study design whilst another study did not examine any table tennis move. The full-text of one article could not be retrieved because it was too old and the journal was closed down [22]. One study was not retrievable with the given digital object identifier (DOI) [23].

The participant characteristics and study designs of the 29 included articles are summarized in Tables 1 and 2, respectively. In brief, participant characteristics, test protocols, and outcome variables of each article were summarized according to playing levels (n = 12), movement tasks (handwork, n = 6; footwork, n = 4; ball/serve against, n = 8) and other factors (n = 4) to identify performance determinants. Six included studies considered multiple factors on different servings with handwork [1,2] or playing level [24,25], racket mass with ball frequency [26], and footwork with footwear [27]. Furthermore, the categorization of dependent and independent variables are mapped in Figure 2. Key findings of the included studies related to playing levels and maneuvers are provided in Tables 3 and 4.

**Figure 2.** A scoping review map summarizing: (**a**) types of forehand and backhand maneuvers; (**b**) types of serves (as variant) to hit back; (**c**) map of dependent variables comparing the number of studies between topics on maneuvers and playing levels; (**d**) body of context (independent variables), the n-values in the interior circle denote number of studies with multiple independent variables between or within the factors stated on the exterior circle; (**e**) direction of strike; and (**f**) shake-hand vs pen-hold.

#### *3.2. Classification of Movement Stage*/*Phase*

While some included studies adopted the maximum or average values of performance outcome of strokes, the majority of the studies divided stroke into movement sub-phases or targeted to selected instants for subsequent analysis. Typically, the stroke was classified into backswing and forward-swing phases, targeted on the specific time points at the termination, backward-end and forward-end [1,3,28–34]. A few included studies [2,24,26,33,35,36] focused on the instant at ball impact which was used to determine the velocity of the racket and ball, while some other included studies investigated the biomechanics at pre-impact and post-impact stages [24,36–38], and over a longer period of time before and after the instant of ball contact [1,2,38,39]. Some included studies endeavored that pelvic and hip rotations were correlated with the racket velocity at impact and thus focused on the starting time of the pelvic forward rotation [36,37]. To sum up, the included studies often investigated the biomechanical parameters at the instant of ball or racket impact as well as the maximum or average value during the time before and after the ball impact.

#### *3.3. Ball and Racket Performance*

Eight included studies examined the effects of ball and racket mechanics as well as serve techniques on table-tennis performance [1,2,24–26,33,35,37], and some of these studies also compared the influences of different handworks [1,2] and playing levels [24,25]. Common variants included the type of ball spin [1,2,33,35,37,39] and the spin rate [24,25]. Moreover, seven included studies investigated ball, racket, and serve as outcome measures instead of variants [31,38,40–45].

Ball speed, accuracy, and repeatability were suggested to be the key indicators of playing level. Ball speed and accuracy were significantly correlated with player ranking in a competition [43]. Higher-level players produced higher ball speed and accuracy, which could be due to significantly shorter duration and variability of duration in the forward swing phase [31,32,38,41]. However, Iino and Kojima [24] found that racket speed at impact was not significantly di fferent between playing levels (advanced vs. intermediate), although players with higher-level can rotate the trunk e ffectively to produce a greater racket acceleration at ball impact. Yet, Iino and Kojima [24] imposed a stringent significance level using a Bonferroni correction. Similarly, Belli et al. [40] found that while there was only a slight di fference in ball speed comparing higher and lower-level players, players with higher-level demonstrated higher accuracy of ball target placement and made fewer errors in training and competition. On the other hand, inexperienced players showed higher inconsistency in ball speed and accuracy during within- and between-day trials [43]. Compared to the intermediate players, advanced players showed smaller variance of joint angle that a ffected the racket vertical angle during forehand topspin stroke [41]. Furthermore, a lower variability in the racket orientation and movement direction could be the reason for more successful returns and higher accuracy of the ball bounce location [38]. An uncontrolled manifold analysis suggested that higher-level players exploited higher degree of redundancy to maintain a similar racket angle at ball impact [41]. In brief, higher-level players exhibited higher accuracy and reproducibility on ball and racket mechanics but may not necessarily produce higher ball speed than lower-level players.

Compared to the topspin serves, returning backspin serves demonstrated significantly higher resultant and vertical racket velocities at ball impact [35,37], which could be contributed greatly by the wrist extension [35]. A possible explanation for this is that backspin serves tend to be treated back-low owing to the spin, resulting in a greater upward velocity of the shoulder joint center [37]. Moreover, peak shoulder torques in all directions, as well as elbow valgus torques, were significantly larger against backspin, in addition to the peaks of upper trunk right axial rotation and extension velocities [37]. Returning a spinning ball also alters the moving distance and velocity of the racket in the upward-downward direction, as compared to an ordinary stroke or a stroke with higher power. Hitting back a backspin serve could be more demanding than a topspin serve.

In addition, biomechanical di fferences between returning light and heavy backspin serves were assessed by two included articles from the same research group [24,25]. They produced di fferent rates of ball backspin (11.4 vs. 36.8 revolutions/s) for light and heavy spin conditions. The heavy spin would direct the racket face to be more open [24]. Furthermore, their results found higher maximum loading at elbow and shoulder joints which might result in higher work done at the racket arm [25]. However, higher-level players showed a higher amount of energy transfer of the elbow for a light spin compared to intermediate players, but the opposite was true for the heavy spin [25], implying significant interaction e ffect between ball spin and playing level. The influence of racket mass and ball frequency were investigated by Iino and Kojima [26], who suggested that a heavier racket could impose higher demand on wrist extension torque, but did not influence trunk and racket arm kinematics and kinetics. A frequent ball serve could result in a lower racket speed at impact possibly since the pelvis and upper trunk rotations were not responsive enough. Table tennis players managed to identify the di fferences in ball spin, frequency, and mass, and accommodated by tilting the racket face angle and adjusting the power output of upper extremity.


#### **Table 1.** Participant characteristics of reviewed studies.


#### **Table 1.** *Cont*.

\* The names of the level or group are adopted from the included studies. Numbers in brackets denote standard deviation. M: male; F: female; Number in bracket denotes standard deviation. NS: not specified; yExp: year of experience; Div: division; h: hours; hrWT: hours per week training; TT: table tennis; Prof.: professionals; w/: with; w/o: without.

#### *3.4. Upper Limb Biomechanics*

There were eight included studies targeting handwork as the variant, while two of them co-variated with different serves (Table 2). Higher racket speed and faster ball rotation were the key attributes of attacking shots and this could be determined by the kinematics/kinetics of upper extremity as well as the efficiency of energy transfer through the upper arm [25,49]. Higher-level players showed significantly larger maximum shoulder internal rotation, elbow varus, and wrist radial deviation torques, in addition to the maximum joint torque power at shoulder joint in both internal and external rotation directions [25]. Higher angular velocity of the wrist joint contributed to a higher ball and

racket speed during drop shot services, while that also produced higher racket speed during long shot services [44].

Moreover, higher-level players rotated the lower trunk efficiently contributing to higher racket speed at ball impact [24]. Meanwhile, the racket horizontal velocity at ball impact was related to the hip axial rotation torque at the playing side (i.e., racket side), while the racket vertical velocity was correlated with backward tilt torques and upward hip joint forces [36]. In contrast, players with shoulder impingement syndrome had sub-optimal coordination and movement patterns of the shoulder girdle [47]. These players significantly reduced muscle activity of the serratus anterior and supraspinatus, which was compensated by increasing overall muscle activity and early activation of upper trapezius [47]. Whole-body coordination and movement would play an important role in driving a speedy ball impact.

Comparing forehand and backhand strokes, racket speed during ball impact was similar but presented differences in the upward and forward velocity components [1]. Forehand stroke lasts slightly longer duration for whole movement cycle and individual phases, and noticeably longer total traveling distance of the racket. This could be because forehand had greater body involvement while the arm and trunk range of motion (RoM) in backhand stroke is limited. Forehand stroke may produce more energy, whilst a longer backswing phase in the high-force condition may generate higher force and longer contact time with the balls [1]. The racket velocity produced by forehand and backhand strokes could be different. During forehand stroke, racket velocity was correlated with the angular velocities of internal arm rotation and shoulder adduction, whereas the racket velocity was correlated with the angular velocities of arm abduction and shoulder rotation during a backhand stroke [2].

A longline forehand topspin produced larger ball rotation, compared to the crosscourt topspin shot. At the instant of the maximum velocity of racket in a forehand topspin stroke, players put their racket more inclined whilst maintaining a more flexed knee and elbow posture, in addition to a more pronounced trunk rotation [45]. Other maneuvers including loop, flick, fast break, and curling ball were also studied [35,42,46]. Compared to curving balls, Zhou et al. [42] suggested that fast breaking significantly reduced racket speed during ball impact. While the flick maneuver was specified as an attack when the ball is closed to the net, there were no detailed explanations on the moves of the fast break and curling ball in which we believed that they could be the flick/drop shot and topspin/sidespin loop maneuvers, respectively. On the other hand, Le Mansec et al. [46] demonstrated that aggressive strokes required greater muscle activities. During smash, biceps femoris, gluteus maximus, gastrocnemius, and soleus muscles were highly activated. Forehand topspin with more power or spin produced significantly higher muscle activation of biceps femoris and gluteus maximus muscles compared to other maneuvers, including backhand top, forehand smash, and flick.

#### *3.5. Lower Limb Biomechanics*

Four included studies investigated different footwork targeting side versus cross-step [4], long versus short chasse step [48], stepping directions and friction [27], and squatting [30], as shown in Tables 1 and 4, while one study compared players of different levels performing a cross-step [34]. Lam et al. [4] identified that both side-step and cross-step footwork produced significantly higher ground reaction force, knee flexion angle, knee moment, ankle inversion and moment compared with one-step footwork, in addition to a significant higher peak pressure on the total foot, toe, first, second and fifth metatarsal regions. On the other hand, long and short chasse steps during a forehand topspin stroke were compared [48]. Long chasse steps produced an earlier muscle activation for vastus medialis, quicker angular velocity, and larger ankle and hip transverse RoM, whereas larger ankle coronal RoM and hip sagittal RoM compared with the short chasse steps [48]. A stable lower limb support base is another important attribute to tackle serve. Yu et al. [30] compared a squat serve with stand serve and found that squat serve produced larger angles and velocities of hip flexion, adduction, knee flexion, and external rotation and ankle dorsiflexion, whereas standing serve produced a higher force-time integral in the rearfoot region. Different stepping angle and footwear friction could also

influence the center of mass and kinematics of knee joint, respectively [27]. Different footwork imposed different lower limb kinematics requirements for table tennis players.




**Table 2.** *Cont*.

NS: not specified; CoM: centre of Mass; w/: with; w/o: without; PP: plantar pressure distribution; EMG: electromyography.

Comparing the lower limb biomechanics among players with various playing levels, Qian et al. [28] and Wang et al. [29] reported distinct findings for respective forehand and backhand crosscourt loops. When executing forehand topspin loop, higher-level players increased knee external rotation, hip flexion and decreased ankle dorsiflexion during backward end phase, and increased hip extension and internal rotation, decreased ankle and knee internal rotation during forward end phase. There was an overall increase in the ankle sagittal RoM as well as hip sagittal and coronal RoM [28]. When performing backhand crosscourt loop against backspin ball, higher-level players increased ankle dorsiflexion, eversion and external rotation, increased knee flexion and abduction and increased hip flexion, adduction, and external rotation at the beginning of backswing, as well as increased ankle dorsiflexion, knee flexion, reduced hip flexion but increased abduction at the end of swing [29]. During cross-step footwork, higher-level players executed superior foot motor control, as indicated by a smaller RoM of foot joints and higher relative load on the plantar toes, lateral forefoot and rearfoot regions [34]. They also demonstrated smaller forefoot plantarflexion and abduction during cross-step end phase but larger forefoot dorsiflexion and adduction during forward end phase [34]. E ffective coordination of lower limb facilitates better upper body rotation in higher-level players [39].

Ba ´nkosz and Winiarski [33] compared inter- and intra-individual variabilities of kinematic parameters. They reported that both variabilities could be quite high, but players attempted to minimize variability at critical moments, such as the instant of ball impact. Higher inter-individual variability could also imply that the technique of coordination movement is rather individual. Adopting or imitating a particular training regime has to pay more attention.

Plantar pressure was also used to evaluate foot loading among di fferent playing levels. When performing forehand loop during backward end phase, higher-level players displayed larger plantar pressure excursion in the medial-lateral direction but smaller in the anterior-posterior direction, accompanied by increased contact areas at midfoot and rearfoot regions while decreased contact area at lesser toe region [3,28]. During forward end phase, higher-level and intermediate players decreased similarly the plantar pressure excursion in the anterior-posterior direction. The contact areas were increased at midfoot, rearfoot, and forefoot regions while decreased at the hallux region [3,28]. The change of plantar pressure excursion and contact area could reflect the strategy compromising dynamic stability and agility in di fferent directions.


**Table 3.** Key findings of included studies comparing playing levels.

#### **Table 3.** *Cont*.


**Table 3.** *Cont*.



ACR: angular changing rate; AP: anteroposterior; BE: backward-end; EMG: electromyography; FE: forward-end; FTA: right forefoot to hindfoot angle; HTA: right hindfoot to tibia angle; ML: mediolateral; PP: peak pressure; RoM: range of motion; XFA: right hallux to forefoot angle. (–) in negative direction/value. The increase/decrease of (–) refer to the absolute magnitude; ↑: significantly higher/larger/increase; ↓: significantly lower/smaller/decrease.

## **4. Discussion**

There was evidence suggesting that higher-level table tennis players produced higher ball accuracy, performance index, and trial-to-trial repeatability in both training and competition. Meanwhile, it was generally perceived that ball and racket velocities were deterministic to playing level since high velocities make the opponent difficult to return the ball. In particular, the maximum racket speed at the moment of impact was regarded as the most important playing technique [1]. However, the current evidence did not come into a consensus that higher-level players necessarily produce higher ball or racket speed. Shoulder joint seems to play an important role to coordinate an effective stroke, as indicated by the effective use of elbow flexion torque, while the power of wrist joint is important during drop shot or long shot services. On the other hand, lower extremities facilitated momentum generation for increased racket velocity. In fact, leg–hip–trunk kinetics accounted for more than half of the energy and muscle force generation in racket sports [28]. Apart from a shorter period of swinging time, the increase in hip flexion and knee external rotations for higher-level players would potentially facilitate a more efficient muscle output to maximize racket velocity through the kinetic chain [28,29], in addition to larger hip and ankle angular velocities [28] which could be correlated with an increased ball speed after ball impact [50]. It should be noted that body coordination movement varies across individuals and trials but players attempted to reproduce movement during critical instants [33]. This was known as functional variability such that players could adapt to the conditions and requirements of the tasks and compensated for the changes with other movement parameters [51]. An optimal training model of body movement could be different among athletes.

Techniques in footwork could play an important role in compromising between dynamic stability and agility to recover back to the ready position for next moves or strokes. Less experienced players tended to have a larger peak ankle dorsiflexion and anterior center of pressure but lesser contact area, which indicated a poorer support base and stability [3,28]. Additionally, a shorter center of pressure in the anterior–posterior direction in higher level players facilitates quicker responses to resume to a neutral position for the next move [3,28]. However, it should be noted that higher level players exhibited larger ankle RoM during the match which may inherit the risk of ankle sprain [28,29].

Regarding the methodological quality, more than half of the included studies did not reveal clearly the source of population and sampling method. There was also a lack of blinding. Although blinding the maneuver conditions seemed to be impossible since the participants needed to be acknowledged for the tasks they performed, it could be accomplished by counting successive returns from delivering random serves by the coaches or serving robots [30,40]. Furthermore, the implementation of a randomized cross-over design across various interventions and maneuvers is necessary to avoid carry-over e ffects. Future studies can investigate how technologies can improve training outcomes. For instance, augmented reality (AR) technology with di fferent filmed footages of di fferent balls and gaze information can be modulated with artificial intelligence program to simulate the virtual opponent with the matched playing levels. Such simulations would provide a steppingstone towards individualized training solutions. On the other hand, several studies investigated a large number of outcome variables which was not well justified. While a full biomechanical profile with a large number of outcome variables were endeavored, statistical analyses were performed without corrections for multiple or multivariate comparisons. This may fall into the trap of data dredging or p-hacking [52] and those research may confine to exploratory studies [53].

There are some limitations when interpreting our findings. A systematic scoping review covered a vast volume of literature over a topic and thus o ffered an overview picture within the discipline [21]. However, due to the heterogeneity and breadth of the included studies, the established data framework did not attempt to answer a single research question which shall be put forward by a systematic review. It is also not possible to conduct meta-analysis to estimate overall determinants on playing levels, movement tasks, and equipment because of the diversity of objectives and designs across the included studies. In fact, the amount of literature required for a subset study was insu fficient to formulate a focused research question for a traditional systematic review. For example, only two included studies were comparing upper limb kinematics of forehand topspin among di fferent player levels in our review. In other words, it is pragmatically demanding to call for more research to establish the map over biomechanical variables, maneuvers, and playing levels, and reinforce key ideas on the determinants of performance using a unified study design and protocol.

Additionally, there was potential publication or language bias since some relevant articles were excluded for being published in Chinese, despite the fact that China is one of the dominating countries in the table tennis sports [16]. Summarizing information from the Chinese literature can enhance the impact of table tennis research but may require considerable e ffort in screening, translation, and interpretation. Furthermore, we found that there was a lack of literature on backspin maneuvers, longline maneuvers, strikes against sidespin ball, and pen-hold players that warrant further investigations.


#### **Table 4.** Key findings of included studies comparing different movement tasks.


#### **Table 4.** *Cont*.


#### **Table 4.** *Cont*.

ACR: angular changing rate; AP: anterior-posterior; BE: backward-end; CoM: center of mass; EMG: electromyography; FE: forward-end; hGRF: horizontal ground reaction force; hV: horizontal velocity; MMV: maximum velocity of the racket; MT: metatarsal; PP: peak pressure; RoM: Range of Motion; RP: ready position; T1: take-off; vGRF: vertical ground reaction force; vV: vertical velocity. (–) in negative direction/value. ↑: significantly higher/larger/increase; ↓: significantly lower/smaller/decrease of the absolute magnitude. # Only highlighted key findings were summarized in the table since these studies included too many outcome variables and/or pairwise comparison results to be listed.
