Background <p>Patellar tendinopathy (PT) is a common overuse condition in sport, but the comparative effectiveness of exercise-based rehabilitation strategies remains uncertain. Earlier syntheses often relied heavily on ranking metrics and included clinically heterogeneous intervention nodes. This study aimed to compare exercise interventions for PT while improving the transparency of intervention classification, transitivity assessment, and sensitivity analyses.</p> Methods <p>This systematic review and network meta-analysis followed PRISMA-NMA guidance and was prospectively registered in PROSPERO (CRD420261296478). Eleven electronic databases were searched from inception to December 2025, and ClinicalTrials.gov was also screened for ongoing or unpublished trials. Randomized controlled trials evaluating exercise-based interventions for clinically and/or imaging-diagnosed patellar tendinopathy were eligible. The primary outcome was the Victorian Institute of Sport Assessment for Patellar Tendinopathy (VISA-P) score. Risk of bias was assessed using the Cochrane RoB 2.0 tool. Because the revised primary network was sparse and several comparisons were informed by limited evidence, the findings were interpreted cautiously. For the revised primary analysis, a conceptually cleaner connected network was constructed by excluding combined exercise-plus-adjunct interventions and one isolated 4-week comparison that did not connect to the main network. A random-effects frequentist network meta-analysis was conducted in R using the netmeta package, with heavy slow resistance training (HSR) as the reference treatment. Additional descriptive and sensitivity analyses were undertaken to examine effect-modifier distributions, follow-up window, and population heterogeneity.</p> Results <p>Seventeen randomized controlled trials were included in the qualitative synthesis. The broader quantitative evidence framework comprised 14 studies (456 participants), which was retained as a sensitivity analysis. The revised primary connected network included 10 studies with 313 participants and 11 treatment nodes. Most trials were judged as having some concerns for overall risk of bias, particularly in domains related to deviations from intended interventions, where adherence, co-interventions, and the practical difficulty of blinding in exercise trials required careful consideration; no trial was judged as having overall high risk of bias. In the primary network, no intervention demonstrated statistical superiority over HSR, and the observed between-intervention differences did not indicate a clinically meaningful hierarchy. Eccentric overload showed the most favorable point estimate versus HSR (SMD 0.33, 95% CI -0.87 to 1.52), followed by blood flow restriction training (0.26, -0.40 to 0.92) and moderate slow resistance (0.12, -0.49 to 0.72). Flywheel training (-1.00, -1.65 to -0.35), eccentric step training (-1.30, -2.59 to -0.02), and concentric exercise (-2.59, -4.16 to -1.01) were estimated to perform less favorably than HSR. Probability-based rankings were exploratory and should not be interpreted as evidence of definitive or clinically meaningful treatment superiority.</p> Conclusions <p>The revised evidence does not support clear superiority of any exercise intervention over HSR for short-term VISA-P improvement in PT, and the observed between-intervention differences do not indicate a clinically meaningful hierarchy. HSR remains a reasonable reference treatment, while several progressive loading-based strategies appear broadly comparable within the limits of the current evidence. Treatment rankings should not be interpreted as definitive, and future trials should use clearer intervention definitions, more comparable populations, and more consistent follow-up schedules.</p>