The influence of basketball players’ tracking speed ability on sports decision performance

Participants

The study used G*Power 3.1.9.7 software to estimate the sample size, setting the effect size to be biased by η² = 0.03 (Lakens, 2014), α = 0.05. The calculation indicated that a statistical test power of 0.80 could be achieved with 40 participants (20 in each group). Considering the potential for participant withdrawal during the experiment, 48 participants were ultimately selected, with 24 participants in each group (Jin et al., 2020). The participants were divided into two groups: an expert group and a novice group, based on their basketball experience and skill level. The expert group consisted of female players from the Northeast Division of the Chinese University Basketball League First Division (Pylyshyn & Annan, 2006; Gou & Li, 2023), with an average playing time of over 15 min per game, a minimum sports level of Level 1 (including Level 1), an average age of (21.20 ± 2.12) years, an average training period of (9.10 ± 2.14) years, and a weekly training time of (27.64 ± 7.16) hours in the past year. The novice group included female students from the basketball elective course at Northwest Normal University, with an average age of (19.83 ± 0.89) years, an average training period of (1.60 ± 0.49) years, a weekly training time of (1.20 ± 0.15) hours, and no formal sports level. All participants have no previous experience in MOT, were right-handed, with normal or corrected vision, and maintained stable emotions and a good mental state before the experiment. They played electronic games for no more than 4 h per week and experienced no fatigue. The experiment was approved by the ethics committee of Northwest Normal University (No. NWNU-20230301). Prior to the experiment, participants were thoroughly informed of the purpose and procedures of the study and were required to sign an informed consent form.

Design

Experiment 1 used a multiple object tracking task to assess tracking performance at different target speeds. A mixed experimental design was employed: 2 (group: experts, novices) × 3 (target speed: 5°/s, 10°/s, 15°/s), where the group was the between-subjects variable, the target speed was the within-subjects variable, and the dependent variable was tracking accuracy (proportion of correctly identified targets). Although the MOT task employed fixed speeds rather than an adaptive procedure, this design choice was made to ensure comparability with previous studies (Hyönä, Li & Oksama, 2019) and to maintain consistency across participants. We acknowledge this as a methodological limitation.

Experiment 2 followed a 2 (group: experts, novices) × 2 (decision type: intuition, cognition) × 3 (attack method: passing, shooting, breakthrough) × 3 (speed: low speed 0.67–3.98 m/s, medium speed 3.99–7.97 m/s, high speed 7.98–12.62 m/s) design. The group was the between-subjects variable, the target speed was the within-subjects variable, and the dependent variable was decision accuracy. Operationally, “intuitive decision” was defined as a rapid response within 600 ms, requiring participants to rely on immediate pattern recognition with minimal deliberation. “Cognitive decision” was defined as a deliberate response within 1,200 ms, requiring participants to integrate situational cues and consciously evaluate options. These definitions are consistent with dual-process models of decision-making (Lu, 2018) and provide an empirical framework for the study. During the experiment, a principal investigator was present to accompany and ensure the smooth progression of the procedure.

Apparatus

The MOT task was presented on a Lenovo E15 laptop, with an operating system of Windows 10, a 15.6-inch display screen, a screen resolution of 1,920 × 1,080 pixels, and a refresh rate of 60 Hz. At the beginning of Experiment 1, a “+” symbol lasting 500 ms was displayed at the center of the screen, followed by 12 white spheres. Three of the spheres turned blue and flashed three times, marking them as target objects, while the remaining spheres remained white as non-target objects. Subsequently, all spheres turned white and moved randomly at speeds of 5°/s, 10°/s, and 15°/s. After 8 s, the spheres stopped moving, and the target objects turned red. Participants were required to determine how many of the red spheres contained target objects, press the corresponding number keys, and proceed to the next trial.

Stimuli

The offensive video sequences used in Experiment 2 were selected from Women’s Chinese Basketball Association (WCBA) games, specifically focusing on the forward position. These clips were drawn from the final matches of the 2020–2021 and 2021–2022 seasons, which were temporally close to the experiment. A total of 129 segments were included, each representing actions such as passing, breakthrough, and shooting (Jin, Ge & Fan, 2023). Ten high-level female basketball players recreated these scenes on-site, with each team wearing black and white jerseys and identified by their numbers. A tall player with short hair wore a sports camera (model: Insta360 One X) to capture footage from the player’s first-person perspective (Ping, 2019).

The decision-making stimuli were video clips derived from WCBA games and subsequently re-enacted by trained female players from a first-person perspective. A total of 129 validated clips represented three offensive actions (passing, shooting, breakthrough) across different speeds.

Validity evidence. To establish content validity, five experienced CUBA coaches and CBA players reviewed all video clips. Structural validity was tested by comparing performance between six high-level players and six university-level players; clips with insufficient discriminative capacity were modified or excluded. Reliability testing indicated acceptable test–retest stability over a two-week interval (r = 0.82 for decision accuracy). For the MOT task, internal consistency was acceptable (Cronbach’s α = 0.85 across blocks).

Video analysis

In the statistical analysis of 60 basketball game videos from Experiment 2, it was noted that there were 10 players in 59 videos, while only one video featured nine players. Due to this fixed number characteristic, analysis of the number of players did not provide effective insights into the results of sports decision-making. Therefore, following consultations with experts, Experiment 2 focused on analyzing the average speed of the ball handler.

Video analysis technology is a critical tool in modern sports science research, providing detailed data on player performance. By recording the time and location (x and y-axis coordinates) of movements during competitions, it is possible to describe the behavior of players through ordinary video footage (Kamble, Keskar & Bhurchandi, 2019). The analysis tracks location changes over a specified time period, which are determined by the nature of the event. Based on the standard dimensions of a basketball court (28 m in length and 15 m in width), the court is set within a Cartesian coordinate system with an x-axis of 28 m and a y-axis of 15 m, with a stride value of 1 m. Each time the step size increases by 1 m, this serves as a reference for the grid covering the field, where mapping the event onto the grid results in corresponding x and y values (Cullinane, Davies & O’Donoghue, 2024). See Fig. 1.

Storm player software (version 5.81) is employed to play the video material and calculate the ball handler’s speed using the following steps:

The time interval is set to Δt (0.5 s), and the position changes of the ball handler are checked within each 0.5-s interval. If there is no change in the player’s position, the location is recorded at 0.5-s intervals. If a change in position occurs, the location is recalibrated based on the time of the change, and the time interval is recalculated. This algorithm enables real-time detection and tracking of the ball handler’s position and movement.

By recording the position changes over a continuous period, the player’s running speed is calculated using the following formula. If the player moves from position P1 (x₁, y₁) to position P2 (x₂, y₂) within time Δt, the velocity v can be expressed as: $v={\displaystyle \frac{P2-P1}{\mathrm{\Delta }t}={\displaystyle \frac{\sqrt{{\left(x2-x1\right)}^2+{\left(y2-y1\right)}^2}}{\mathrm{\Delta }t}.}}$

By calculating the average speed of the ball handler in each video and combining it with the speed settings of the multiple object tracking task in Experiment 1, the decision video material was divided into three speed ranges: 0.67–3.98, 3.99–7.97, and 7.98–12.62 m/s. The results showed that, within the range of 0.67–3.98 m/s, there were a total of 28 videos (including six passes, 13 shots, and nine breakthroughs). In the range of 3.99–7.97 m/s, there were a total of 27 videos (including 11 passes, seven shots, and nine breakthroughs). In the range of 7.98–12.62 m/s, there were only five videos (including three passes and two breakthroughs), with no shooting decisions made by the ball handler while running at high speed, which constitutes a limitation in stimulus balance and may affect cross-condition comparisons.

Procedure

Experiment 1: A multiple object tracking experimental program was developed using Matlab R2020b software. The distance between the participant and the screen was approximately 60 cm, and the stimulus presentation area covered the entire screen. To familiarize the participants with the experimental process, five practice trials were conducted before the formal experiment (Sagehorn et al., 2024). The experiment consisted of 10 blocks, each containing three trials corresponding to the three target speeds: 5°/s, 10°/s, and 15°/s, for a total of 30 trials. A 10-s white screen was displayed between each block to alleviate eye fatigue. The experimental sequence was counterbalanced within subjects. The entire experiment lasted approximately 11 min.

Experiment 2 was programmed using Experiment Builder 2.3 software. Referring to previous research, the presentation time for intuitive decisions was set at 600 ms, while the presentation time for cognitive decisions was set at 1,200 ms (Lu, 2018). Intuitive decision-making task instructions: “The ‘+’ sign is first displayed in the center of the screen, followed by a blank screen and a first-person perspective game video, prompting: ‘Please make a decision.’ Carefully observe and judge the attacking style of the ball handler within 0.6 s. The situation on the court is urgent; please make a quick decision. Press 1 for passing, press 2 for shooting, and press 3 for a breakthrough.” Cognitive decision-making task instructions: “The ‘+’ sign is first displayed in the center of the screen, followed by a blank screen and a first-person perspective game video, prompting: ‘Please make a decision.’ Carefully observe and judge the attacking style of the ball handler within 1.2 s. Please make accurate and prompt decisions. Press 1 for passing, press 2 for shooting, and press 3 for a breakthrough.” After the practice trials are completed, the screen will display “Start formal experiment,” and the button operation will remain consistent with the preparation stage. The video will be played in full screen, and participants will alternate between cognitive and intuitive decision-making tasks using an ABBA design to balance task order (Liu, 2012), with a 30-s break between tasks. Three types of attack videos for each decision-making task will be randomly presented, and the entire experimental process will take approximately 27 min. See Fig. 2.

Statistical analysis

Data from Experiment 1 (tracking accuracy) and Experiment 2 (decision accuracy) were analyzed with repeated measures ANOVAs. Normality of residuals was confirmed using the Shapiro–Wilk test prior to conducting ANOVAs. Where interactions were significant, simple effects analyses with Bonferroni correction were performed. Statistical significance was set at P < 0.05.

Decision accuracy of experts and novices at different video speeds

A four-way repeated measures ANOVA was conducted with group (experts, novices) as the between-subjects factor, and decision type (intuition, cognition), attack method (passing, shooting, breakthrough), and speed (low: 0.67–3.98 m/s, medium: 3.99–7.97 m/s, high: 7.98–12.62 m/s) as within-subjects factors, using decision accuracy as the dependent variable. Results are presented in Table 1.

Main effects

Experts (0.66 ± 0.11) performed significantly better than novices (0.32 ± 0.10), F(1, 46) = 1,069.644, P < 0.001, η² = 0.959, d = 3.20.

Cognitive decisions (0.52 ± 0.10) were more accurate than intuitive decisions (0.47 ± 0.11), F(1, 46) = 45.818, P < 0.001, η² = 0.499, d = 0.48.

For attack method, accuracy differed significantly, F(2, 92) = 171.733, P < 0.001, η² = 0.789. Shooting (0.40 ± 0.10) was significantly less accurate than passing (0.55 ± 0.11, d = 1.36) and breakthrough (0.53 ± 0.12, d = 1.20), while passing and breakthrough did not differ significantly (P = 0.05).

Speed had a strong effect, F(2, 92) = 729.000, P < 0.001, η² = 0.941, with accuracy decreasing across low, medium, and high speeds (Ps < 0.001, ds = 0.94–1.58).

Two-way interactions

Group × decision type: experts outperformed novices in both intuitive and cognitive tasks (Ps < 0.001, ds = 2.8–3.1). In both groups, cognitive accuracy exceeded intuitive accuracy (Ps < 0.01).

Group × attack method: across passing, shooting, and breakthrough, experts outperformed novices (Ps < 0.001, ds = 2.5–3.3). In experts, shooting accuracy (0.52 ± 0.10) was lower than passing (0.74 ± 0.12, d = 1.98) and breakthrough (0.72 ± 0.13, d = 1.80). Similar patterns were found in novices.

Group × speed: experts outperformed novices at all three speeds (Ps < 0.001, ds = 2.6–3.4). In both groups, accuracy decreased progressively as speed increased (Ps < 0.001).

Decision type × attack method: cognitive accuracy was higher than intuitive across all attack methods (Ps < 0.05), except passing vs breakthrough where no difference was found.

Higher-order interactions

Significant three-way and four-way interactions were observed (see Table 1). For instance, in expert cognitive decisions, passing (0.78 ± 0.12) and breakthrough (0.77 ± 0.11) accuracy was significantly higher than shooting (0.52 ± 0.10, Ps < 0.001, ds = 2.1–2.3). In novices, accuracy was uniformly lower, but the relative disadvantage for shooting remained consistent.

Overall, these analyses highlight that expert advantage was robust across all decision contexts, with effect sizes consistently large (Cohen’s d > 0.8), whereas task difficulty (speed, shooting scenarios, intuitive conditions) disproportionately impaired novice performance.

Group × speed

A 2 (group: expert, novice) × 3 (speed) repeated-measures ANOVA revealed a significant interaction, F(2, 118) = 42.37, P < 0.001, η² = 0.42. Simple-effects analyses showed that experts outperformed novices at all speeds (Ps < 0.001; Cohen’s d = 0.96–1.22). For both groups, decision accuracy declined as speed increased (Ps < 0.001; linear trend η² = 0.38).

Decision type × attack method

The Decision Type × Attack Method interaction was significant, F(2, 118) = 8.15, P < 0.001, η² = 0.12. Cognitive decisions were more accurate than intuitive ones across attack methods (Ps < 0.05; d = 0.31–0.44). Shooting accuracy was lower than passing and breakthrough (Ps < 0.001; d = 0.55–0.78); passing vs. breakthrough did not differ (P > 0.05).

Decision type × speed

A significant interaction emerged, F(2, 118) = 10.29, P < 0.001, η² = 0.15. At low speed (0.67–3.98 m/s) cognitive and intuitive accuracy did not differ (P = 0.108). At medium and high speeds cognitive accuracy exceeded intuitive accuracy (Ps < 0.001; d = 0.48–0.63).

Attack method × speed

Attack method interacted with speed, F(4, 236) = 6.74, P < 0.001, η² = 0.10. At 7.98–12.62 m/s, shooting accuracy was lower than passing and breakthrough (Ps < 0.001; d = 0.52–0.70); no other pairwise differences were significant.

Higher-order interactions

Group × Decision Type × Attack Method: Experts showed higher accuracy than novices across all combinations (Ps < 0.001; d = 0.82–1.19). Within experts, cognitive accuracy exceeded intuitive accuracy for passing and breakthrough (Ps < 0.001; d = 0.40–0.58). Novices showed this advantage only for passing (P < 0.01; d = 0.34).

Group × Decision Type × Speed: Experts outperformed novices at all speeds (Ps < 0.001; d = 0.88–1.15). Cognitive accuracy surpassed intuitive accuracy for medium and high speeds (Ps < 0.05; d = 0.36–0.52).

Group × Attack Method × Speed: Experts were more accurate across all attack methods and speeds (Ps < 0.001; d = 0.80–1.07). Shooting accuracy declined significantly with increasing speed (Ps < 0.05; η² = 0.09).

Decision Type × Attack Method × Speed: Cognitive accuracy was higher at high speed for passing, at medium speed for shooting, and at medium/high speed for breakthrough (Ps < 0.05; d = 0.29–0.47). Shooting accuracy at high speed was again lowest across methods (Ps < 0.001; d = 0.51–0.73).