Purpose: To compare the accuracy in the estimation of the Smith machine bench press 1-repetition maximum (1RM) when using a novel minimum velocity threshold (MVT) called optimal MVT (MVT that minimizes the differences between the actual and predicted 1RM in a preliminary session) with respect to using the 2 standard MVTs (general and individual MVTs). Methods: A total of 126 young men (Smith machine bench press 1RM = 80.7 [13.6] kg) completed 2 identical sessions consisting of an incremental loading test until reaching the 1RM load. Four individual load–velocity relationships were modeled in each session considering all loading conditions until reaching the load that showed the closest mean velocity to 0.60, 0.50, 0.40, and 0.30 m·s−1. The first testing session was used to determine the preindividual MVT and 4 optimal MVTs (1 for each final test velocity), while the second testing session was used to estimate the 1RM using 4 types of MVT (general MVT, preindividual MVT, actual-individual MVT, and optimal MVT). Results: The absolute errors in the prediction of the 1RM were significantly lower for the optimal MVT (2.94 [2.40] kg) compared to the general MVT (3.66 [2.99] kg), preindividual MVT (3.80 [3.15] kg), and actual-individual MVT (4.02 [3.21] kg). The optimal MVT (intraclass correlation coefficient [ICC] ranged from .56 to .62) was always more reliable than the individual MVT (ICC = .34). Conclusions: The optimal MVT provides more accurate estimates of the Smith machine bench press 1RM than the standard MVTs previously used in scientific research (general and individual MVTs).