*2.1. Participants*

Twenty (*n* = 20) male basketball players were recruited from local universities. Their average age, height and body mass were 22.6±1.1 years, 179.4±3.4 cm and 72.7±8.2 kg, respectively. All participants had at least 4 years of experiences in playing basketball and were right hand dominant single-handed shooters. The average basketball training experience and training time were 8.5 ± 2.4 years and 5.2 ± 1.6 h per week, respectively. All participants were physically fit and healthy and reported no injuries over the previous 6 months. Ethical approval (IRB-2017-BM-006) was granted from the institutional ethics committee. Written informed consent was obtained from all participants.

#### *2.2. Experimental Conditions and Procedure*

All free-throw shooting conditions were performed in our biomechanical laboratory. The freethrow distance and the height of the basketball rim were set according to the International Basketball Federation standards [19]. The participants performed single-handed free-throws under five di fferent garmen<sup>t</sup> conditions, control-pre: no garmen<sup>t</sup> pre-control, Top: upper-body compression garmen<sup>t</sup> (Li Ning, Powershell, AULM043-I, Beijing, China), Bottom: lower-body compression garmen<sup>t</sup> bottom (Li Ning, Powershell, AUDL101-1, Beijing, China), full: both upper-body and lower-body compression garmen<sup>t</sup> and control-post: no garmen<sup>t</sup> post-control, as shown in Figure 1. Control-pre and control-post were the first and the last test conditions. The remaining three compression garmen<sup>t</sup> conditions (top, bottom and full) were randomly assigned as the second to the fourth conditions across participants. As the experimental protocol compared the first and last conditions, we were able to evaluate the fatigue e ffect [22]. For each free-throw condition, 20 free-throw shooting trials were performed. Testing of the next condition started immediately after the participant changed their garments.

**Figure 1.** Compression garmen<sup>t</sup> conditions: (**a**) top; (**b**) bottom; (**c**) full (top + bottom).

The control conditions (control-pre and control-post) were self-selected comfortable sportswear that were not compression garments. The experimenters measured the height, waist and chest circumference of the participants to determine the appropriate garmen<sup>t</sup> [23]. The appropriate compression garmen<sup>t</sup> size was pre-determined by the manufacturer's sizing guidelines and was based on the body height and mass of each participant. Next, we assigned participants compression garments one size smaller than the pre-determined appropriate size in order to increase the interfacial pressure, as recommended by the experimental protocol detailed by Williams and colleagues [12].

A motion capturing system with multiple inertial measurement units (MyoMOTION, Noraxon, Inc., Scottsdale, AZ, USA) was used to measure full-body kinematics during the free-throw shooting trials. The inertial measurement units (IMU) were attached and strapped to each body segmen<sup>t</sup> according to the instrument guidelines. During each free-throw trial, the participants performed shooting from the same position behind the free-throw line. The sampling frequency of the IMU was 200 Hz. The kinematic data during the free-throw motion were post-processed using Matlab software (MathWorks, Inc., Natick, MA, USA) using a 6 Hz cuto ff 4th order Butterworth low-pass filter.

#### *2.3. Outcome Measures*

Outcome measures including performance score (accuracy) and joint ROM variables were investigated. The performance score was gauged using an ordinal six-point (0 to 5 point) scoring system. Five, four and three points denoted a clean score, that the ball hit the rim and went in, and that the ball hit the backboard and went in, respectively. Two, one and zero points denoted that the ball hit the rim and missed, hit the backboard and missed and missed complete, respectively, as illustrated in Table 1 [19,24]. The consistency of the score was also assessed by the coe fficient of variation (i.e., the ratio of the standard deviation to the mean of the trials).


**Table 1.** The six-point basketball shooting performance score system.

ROM of the head, trunk, elbow, shoulder, wrist, hip, knee and ankle joints in the sagittal, coronal and frontal planes were calculated. Data were averaged across trials for each participant in each condition which served as the targeted average profile for subsequent statistical analysis [25]. We did not view the within-participant e ffect (trial) of ROM as an independent observation or random factor to be analyzed.

#### *2.4. Data Analysis*

All statistical analysis was performed in SPSS 21 (IBM, New York, NY, USA). Prior to statistical analysis, the Shapiro–Wilk test was performed to check for the normality of the kinematic data, and it was satisfied. The Wilcoxon signed-rank test was performed to compare free-throw performance scores between the control-pre- and control-post-control conditions to ensure that there was no learning or fatigue e ffect (i.e., Control pre- and post-control were not significantly di fferent). Furthermore, one-way repeated measures analysis of variance (ANOVA) was performed to examine any significant di fference for joint ROM variables between the control-pre, top, bottom and full conditions, followed by the post hoc pairwise comparison of Least Significant Di fference (LSD) if a significant main e ffect was found. We chose the LSD approach as our research hypothesis was more focused on planned comparisons. As such, we regarded the ANOVA as an additional constraint [26]. Similarly, the comparison for the performance score and the coe fficient of variation was performed using a nonparametric test (Friedman test), with the post hoc pairwise Wilcoxon signed-rank test, as the performance score was gauged in an ordinal scale. Level of significance was set at *p* = 0.05. The indices of e ffect size for the ANOVA and post hoc pairwise comparison were partial η2 and Cohen's d, respectively.
