Background: This study determined the accuracy of different velocity-based methods when predicting one-repetition maximum (1RM) in young and middle-aged resistance-trained males. Methods: Two days after maximal strength testing, 20 young (age 21.0 ± 1.6 years) and 20 middle-aged (age 42.6 ± 6.7 years) resistance-trained males completed three repetitions of bench press, back squat, and bent-over-row at loads corresponding to 20–80% 1RM. Using reference minimum velocity threshold (MVT) values, the 1RM was estimated from the load-velocity relationships through multiple (20, 30, 40, 50, 60, 70, and 80% 1RM), two-point (20 and 80% 1RM), high-load (60 and 80% 1RM) and low-load (20 and 40% 1RM) methods for each group. Results: Despite most prediction methods demonstrating acceptable correlations (r = 0.55 to 0.96), the absolute errors for young and middle-aged groups were generally moderate to high for bench press (absolute errors = 8.2 to 14.2% and 8.6 to 20.4%, respectively) and bent-over-row (absolute error = 14.9 to 19.9% and 8.6 to 18.2%, respectively). For squats, the absolute errors were lower in the young group (5.7 to 13.4%) than the middle-aged group (13.2 to 17.0%) but still unacceptable. Conclusion: These findings suggest that reference MVTs cannot accurately predict the 1RM in these populations. Therefore, practitioners need to directly assess 1RM.