Billions of organisms, from bacteria to humans, migrate each year and research on their migration biology is expanding rapidly through ever more sophisticated remote sensing technologies. However, little is known about how migratory performance develops through life for any organism. To date, age variation has been almost systematically simplified into a dichotomous comparison between recently born juveniles at their first migration versus adults of unknown age. These comparisons have regularly highlighted better migratory performance by adults compared with juveniles, but it is unknown whether such variation is gradual or abrupt and whether it is driven by improvements within the individual, by selective mortality of poor performers, or both. Here we exploit the opportunity offered by long-term monitoring of individuals through Global Positioning System (GPS) satellite tracking to combine within-individual and cross-sectional data on 364 migration episodes from 92 individuals of a raptorial bird, aged 1-27 years old. We show that the development of migratory behaviour follows a consistent trajectory, more gradual and prolonged than previously appreciated, and that this is promoted by both individual improvements and selective mortality, mainly operating in early life and during the pre-breeding migration. Individuals of different age used different travelling tactics and varied in their ability to exploit tailwinds or to cope with wind drift. All individuals seemed aligned along a race with their contemporary peers, whose outcome was largely determined by the ability to depart early, affecting their subsequent recruitment, reproduction and survival. Understanding how climate change and human action can affect the migration of younger animals may be the key to managing and forecasting the declines of many threatened migrants.