Resumo. ABSTRACT The aim was to verify the reliability and sensitivity of countermovement jump (CMJ) derived variables in detecting small, moderate and large changes and whether the capacity of CMJ-derived variables in detecting fatigue is dependent of the volume of the fatiguing exercise. Seventeen physically active men performed two fatigue protocols, on separate weeks, composed by continuous vertical jumps: short protocol (7 x 10 jumps) and long protocol (14 x 10 jumps). Jump height (JH), power output (PO), impulse (IMP) and vertical stiffness (KVERT) were measured during CMJ prior to and immediately after the fatigue protocols. Intraclass coefficient correlation, typical error, smallest worthwhile change and magnitude-based inference were analyzed. PO and JH presented excellent reliability and good sensitivity to detect small and medium changes, respectively. Negative effects of fatigue could be detected most likely by PO, regardless of fatiguing exercise volume. JH and IMP seem to be affected only after long protocol and KVERT only after short protocol. In conclusion, PO (peak and mean) is the better marker in CMJ with excellent reliability and sensibility, which allows detect even the small effects and differentiate the fatigue levels.