Tennis requires skill, physical attributes, and strategy. Ball velocity and placement are two of the most important components in winning the faster-paced modern game. Although isokinetic testing has been used to evaluate physical characteristics and injury potential in tennis players, few studies have compared isokinetics and on-court performance. Such a comparison would help establish links between speed-specific properties of functioning muscles and stroke production and could affect overall training strategy. This study compared isokinetic peak torque (PT), average power (AP), and total work (TW) during specific testing patterns correlated with ball velocity or stroke accuracy during the service, forehand, and backhand and developed predictive equations for each stroke using these variables. Thirty-five players, aged 13-18 years with at least 4 years playing experience, were evaluated using internal and external shoulder rotation, leg extension, and diagonal throwing motions. Ball velocity was measured using a radar gun. Accuracy was evaluated on the basis of shot position and depth. Significant correlations were found between ball velocity and a number of isokinetic variables, while no significant correlations were observed with shot accuracy. Significant isokinetic variables for each stroke were entered into regression models. One isokinetic speed sufficiently predicted ball velocity for each stroke, since no increase in predictive capacity was observed with the addition of other isokinetic parameters. We conclude that isokinetics at testing speeds between 1.57 and 4.71 rad.s(-1) can effectively predict ball velocity, but not accuracy, and that our results may be helpful in planning strategies for training and rehabilitation.