This study compared Competency-Based Curriculum (CBC) implementation outcomes in Chemistry and Physics performance among secondary school learners following the post-2020 curriculum reform. Despite the reform's emphasis on competency-based, learner-centred pedagogy and continuous formative assessment, science subjects such as Chemistry and Physics have continued to record uneven performance trends, and it remained unclear whether the reform had translated into comparable achievement gains across the two disciplines or whether disparities in implementation fidelity, teacher preparedness, and school resourcing accounted for observed differences. Using a comparative cross-sectional design, the study drew on a simulated but contextually grounded dataset of 420 learners (211 offsetting Chemistry and 209 offsetting Physics) drawn from a purposive sample of secondary schools, capturing pre-reform baseline scores, post-reform CBC performance scores, a CBC fidelity index, teacher CBC training exposure, a school resource index, and a learner engagement score. Data were analysed using univariate descriptive statistics, bivariate independent-samples t-tests and Pearson correlations, and multivariate ordinary least squares (OLS) regression, with sensitivity analyses conducted using heteroskedasticity-consistent (HC3) robust standard errors, an outlier-adjusted re-estimation, and a subject-by-fidelity interaction specification to test the robustness of the findings. The multivariate model explained approximately 46.6 percent of the variance in post-reform performance scores and showed that CBC fidelity, teacher training exposure, school resourcing, prior achievement, and urban school location were significant positive predictors of post-reform performance, while Physics learners scored marginally but significantly lower than Chemistry learners after controlling for these factors. Bivariate comparison of raw mean scores, however, showed no statistically significant difference between the two subjects, indicating that the subject-level disadvantage in Physics only became detectable once implementation-related covariates were accounted for. Sensitivity analyses confirmed that the fidelity effect was stable in magnitude and significance across robust-