In this study, we explore mechanical constraints on the swimming performance of zebrafish larvae (Danio rerio) that might explain why larvae switch from sustained swimming to the more efficient burst & coast as they grow. Two hypotheses have been proposed to explain why young fish larvae perform poorly at burst & coast. First, their initial momentum might be low; second, their drag coefficient might be high. To test the two hypotheses, this study makes a quantitative comparison between experimental observations of swimming fish larvae and a CFD model of a self-propelled fish. The study focuses on larvae of the crucial age and size range in which zebrafish switch swimming style. Our studies show that hatchlings perform poorly not only because they cannot accelerate to a high initial coasting speed and hence do not gain enough initial momentum. But they also suffer higher decelerations while coasting due to a high drag coefficient. Overall, the fivefold difference in coasting distance between hatchlings and older larvae corresponds closely to a threefold difference in the time constant of the speed decay and a threefold difference in initial momentum. Our data also show that swimming speed does not decay exponentially, as predicted by the drag-speed relationship in the viscous flow regime, but hyperbolically, due to flow phenomena developing in the boundary layer during the coast.