Actually, 10-20 lbs difference in tension can represent up to 7-8 mph difference in ball velocity. How can this be? The 2-3 mph difference typically reported by the tennis science folks represents a measurement taken in the racquet frame of reference. When you transform the ball velocity vectors to the court frame of reference, you need to take into account the effect of the difference in rebound angle. As we know, a lower tension results in a higher rebound angle (for a topspin shot), so the ball rebounds in a direction more closely aligned with the racquet's swingpath (i.e, a lower tension delivers less of a glancing blow and more or a direct momentum transfer). This means that the higher the rebound angle, the more total momentum is being transferred to the ball. Some of this extra momentum goes into extra spin, and the rest of this extra momentum goes into extra ball velocity. But in the racquet frame of reference (in which the measurements are made), this fact gets hidden because the rebound angle's effect on the ball velocity is unaccounted for. To accurately measure the effect of tension on ball speed, you'd need to do a tedious experiment where you held the rebound angle constant via a trial-and-error iterative procedure for each data point. That's not that practical in the lab unless you have a lot of time on your hands. But it is not too hard to get a rough estimate for the true difference in ball speed by taking ball speed and rebound angle data collected in the racquet frame of reference and then transforming it to court frame of reference. You also need to input values for the racquet speed and swingpath. You'd have to neglect the effects of different incoming impact angle (or else look for a way to correct for them), but you'd get a reasonable estimate. I have only done the calculation for a few conditions using data from the tennis science literature, and based on these, the true ball speed difference is usually 2-3x the published difference based on racquet frame of reference measurements alone.