The Modified 3-Minute All-Out Test Parameters as Predictors of 50-, 100-, and 200-m Front Crawl Official Performance in Trained Swimmers

: Di Gennaro, S, Panascì, M, Apollaro, G, Fassone, M, Puce, L, Ruggeri, P, and Faelli, E. The modified 3-minute all-out test parameters as predictors of 50-, 100-, and 200-m front crawl official performance in trained swimmers. J Strength Cond Res XX(X): 000-000, 2026-In this study, we investigated (a) the relationship between the modified 3-minute all-out test (3mAOmod) parameters and swimming official performance in 50-, 100-, and 200-m front crawl; (b) the predictive accuracy of critical speed (CS) model; (c) the predictive linear regression (MLR) models incorporating CS, curvature constant (D') or anaerobic speed reserve (ASR), and anthropometry. Twenty-three young swimmers (mean ± SD: age = 17.5 ± 1.9 years, mass = 61.8 ± 7.2 kg, height 1.79 ± 0.08 m) performed the 3mAOmod and, within 6 weeks, competed in 50-, 100-, and 200-m official races. The relationships between 3mAOmod parameters and race times were examined using Pearson's and Spearman's correlation. Race times were predicted using the CS model and MLR, including CS, D', or ASR, sex, age, height, and body mass. Mean absolute error (MAE) and paired sample t-tests with Bonferroni corrections were used to compare actual and predicted results. Significance was set at p ≤ 0.05. The correlation between CS and race time increased with the increase in race distance (r = -0.491 to -0.698) while the correlation between D' and ASR with race time decreased with the increase in race distance (rho = -0.628 to -0.347 and -0.574 to -0.274, respectively). The CS model underestimates race times in the 50- and 200-m races (p ≤ 0.001, MAE = ±3.33 and ±5.06 seconds, respectively), but not in the 100-m race (p = 0.070, MAE = ±1.72 seconds). Predictions from MLRs were not significantly different from actual performance in all events (p = 0.999, ±0.41 seconds < MAE < ±3.56 seconds). The 3mAOmod parameters are related to 50-, 100-, and 200-m performance, supporting their physiologic rationale and, when applied to the developed MLRs, can predict performance with higher accuracy than the CS model, and in a user-friendly manner.