Purpose: to compare different methods to assess the arm stroke efficiency ( ${\eta }_F$ ), when swimming front crawl using the arms only on the Measurement of Active Drag System (MAD System) and in a free-swimming condition, and to identify biophysical adaptations to swimming on the MAD System and the main biophysical predictors of maximal swimming speed in the 200 m front crawl using the arms only ( $v_{200m}$ ). Methods: fourteen swimmers performed twice a 5 × 200 m incremental trial swimming the front crawl stroke using the arms only, once swimming freely, and once swimming on the MAD System. The total metabolic power was assessed in both conditions. The biomechanical parameters were obtained from video analysis and force data recorded on the MAD System. The ${\eta }_F$ was calculated using: (i) direct measures of mechanical and metabolic power (power-based method); (ii) forward speed/hand speed ratio (speed-based method), and (iii) the simplified paddle-wheel model. Results: both methods to assess ${\eta }_F$ on the MAD System differed (p < 0.001) from the expected values for this condition ( ${\eta }_F$ = 1), with the speed-based method providing the closest values ( ${\eta }_F$ ~0.96). In the free-swimming condition, the power-based ( ${\eta }_F$ ~0.75), speed-based ( ${\eta }_F$ ~0.62), and paddle-wheel ( ${\eta }_F$ ~0.39) efficiencies were significantly different (p < 0.001). Although all methods provided values within the limits of agreement, the speed-based method provided the closest values to the “actual efficiency”. The main biophysical predictors of $v_{200m}$ were included in two models: biomechanical (R² = 0.98) and physiological (R² = 0.98). Conclusions: our results suggest that the speed-based method provides the closest values to the “actual ${\eta }_F$ ” and confirm that swimming performance depends on the balance of biomechanical and bioenergetic parameters