The effectiveness of different training modes on preplanned change-of-direction performance in basketball players: a systematic review and network meta-analysis

To systematically compare the relative effects of different training modalities on preplanned change-of-direction (COD) performance in basketball players using a Bayesian network meta-analysis. Web of Science, PubMed, ProQuest, and EBSCOhost were systematically searched from inception to July 17, 2026, for randomized and non-randomized controlled studies investigating longitudinal training interventions in basketball players. Eligible outcomes were restricted to preplanned, non-ball, single-bout COD performance assessed using timed tests. Risk of bias was evaluated using the Cochrane Risk of Bias 2 tool for randomized controlled trials and ROBINS-I for non-randomized studies. A Bayesian random-effects network meta-analysis was performed using Markov chain Monte Carlo methods. Standardized mean differences (SMDs) with 95% credible intervals (CrIs) were estimated, with negative values indicating effects favoring the active intervention. Global and local inconsistency were assessed using the design-by-treatment interaction test and node-splitting analyses, respectively. Treatment rankings were estimated using the surface under the cumulative ranking curve (SUCRA). Sensitivity analyses, subgroup analyses by sex and age, and exploratory network meta-regressions for mean age and intervention duration were also conducted. Forty-three studies involving 1,203 basketball players and nine training modalities were included. Compared with regular basketball training (BT), combined plyometric and change-of-direction training (PT-COD; SMD = − 1.54, 95% CrI: −2.34 to − 0.75), eccentric training (ET; SMD = − 1.10, 95% CrI: −2.03 to − 0.15), plyometric training (PT; SMD = − 0.76, 95% CrI: −1.10 to − 0.44), and change-of-direction training (CODT; SMD = − 0.72, 95% CrI: −1.23 to − 0.20) significantly improved preplanned COD performance. In contrast, complex training (CT; SMD = − 0.74, 95% CrI: −1.56 to 0.09), small-sided games (SSG; SMD = − 0.59, 95% CrI: −1.44 to 0.25), sprint-interval training (SIT; SMD = − 0.43, 95% CrI: −1.01 to 0.15), and resistance training (RT; SMD = − 0.06, 95% CrI: −0.68 to 0.56) showed no clear advantage over BT. PT-COD achieved the highest SUCRA value (94.50%), followed by ET (76.81%), PT (60.61%), CODT (57.09%), and CT (56.85%). No significant global inconsistency was detected ( p = 0.19), although local inconsistency was identified for ET versus BT and PT-COD versus BT. Sensitivity analyses consistently supported the benefits of PT-COD, ET, and PT, whereas the effect of CODT showed some sensitivity to restriction by COD test type. Exploratory meta-regression found no clear evidence that mean age or intervention duration modified treatment effects. The comparison-adjusted funnel plot and Egger’s regression test ( p = 0.011) suggested potential small-study effects. PT-COD showed the greatest overall potential for improving preplanned COD performance in basketball players, whereas PT was supported by the most consistent evidence across the primary and sensitivity analyses. ET also appeared promising, and CODT produced a beneficial effect in the primary analysis. However, the high rankings of PT-COD and ET should be interpreted cautiously because these interventions were supported by relatively few studies and were affected by local inconsistency. Training programs targeting COD performance should therefore emphasize rapid force production together with braking and directional-change demands, while treatment rankings should be interpreted in conjunction with effect magnitude, uncertainty, and the strength of the underlying evidence.