Velocity-based training is a method used to monitor resistance-training programs based on repetition velocities measured with tracking devices. Since velocity measuring devices can be expensive and impractical, trainee’s perception of velocity (POV) may be used as a possible substitute. Here, 20 resistance-trained males first completed 1RM tests in the bench-press and squat. Then, in three counterbalanced sessions, participants completed four sets of eight repetitions in both exercises using 60%1RM (two-sessions) or 70%1RM. Starting from the second repetition, participants reported their POV of each repetition as a percentage of the first repetition. Accuracy was calculated as the difference between POV and actual-velocity measured with a linear-encoder. Two key findings emerged. First, the absolute error in both exercises was ~5.8 percentage-points in the second repetition, and it increased to 13.2 and 16.7 percentage-points by the eighth repetition, in the bench-press and squat, respectively. Second, participants were 4.2 times more likely to underestimate velocity in the squat. Given the gradual increments in the absolute error, POV may be better suited for sets of fewer repetitions (e.g., 4-5) and wider velocity-loss threshold ranges (e.g., 5-10%). Given the systematic underestimation error in the squat, a correction factor may increase POV accuracy in this exercise.