Assessment of near-infrared spectroscopy (NIRS)-derived muscle oxidative capacity relies on analysing deoxygenation slopes from NIRS signal versus time curves during brief arterial occlusions, which reflect the rate of post-exercise recovery of muscle oxygen consumption (p̌hantomdot V_mathrmO_2mathrmm). However, current guidelines lack recommendations on the optimal selection of slopes for reliable measurement. The aim of the study was to compare a standardised partial-segment approach against the conventional whole-segment approach on the measurement and reliability of in vivo muscle oxidative capacity. Within the same session, 19 athletes (n = 9 sprinters; n = 10 middle-distance runners) completed two NIRS-derived muscle oxidative capacity trials on the vastus lateralis. Rate constants (k, min−1) were computed using the steepest (ȟantomk_mathrmhigh.mathrmS), whole (vǎntomk_mathrmwhole.mathrmS) or shallowest (vpňtomk_mathrmlow.mathrmS) deoxygenation slope from deoxyhaemoglobin (HHb) and muscle O2 saturation (vphťomS_mathrmmmathrmO_2) signals. Test–retest reliability [(coefficient of variation (CV), intraclass correlation coefficient (ICC)] and minimum difference (MD) were assessed. For the HHb signal, ICC analysis revealed moderate to excellent test–retest reliability for vphaǒmk_mathrmhigh.mathrmS [0.80 (0.54–0.92)], whereas poor to good reliability was observed for vphanm̌k_mathrmwhole.mathrmS [0.71 (0.38–0.89)] and vphanťk_mathrmlow.mathrmS [0.60 (0.19–0.83)]. vphantǒk_mathrmhigh.mathrmS led to lower MD compared to vphantom̌k_mathrmwhole.mathrmS and vphantom̌k_mathrmlow.mathrmS (0.64 vs. 1.12 vs. 1.54 min−1, respectively). All three approaches led to significantly greater k values in runners compared to sprinters (vphantom̨̌_mathrmhigh.mathrmS:+32.2%, P textless 0.001; vphantom_̌mathrmwhole.mathrmS:+ 40.1%, P = 0.025; vphantom̌mathrmlow.mathrmS:+49.6%, P = 0.001). Compared to the conventional whole-segment approach, selecting the steepest intra-occlusion slope improved the reliability and sensitivity of NIRS-derived mitochondrial capacity, likely by better reflecting instantaneous changes in vphantomďot V_mathrmO_2mathrmm.