We report a laboratory experiment on dynamic patent races in an indefinite horizon with complete information. In a competitive environment, we analyze the R&D investment behavior of players who are randomly and anonymously paired in each race. We vary subjects’ initial positions as leader/follower or symmetric/asymmetric, as well as the distance between the initial knowledge stock and the target. Our results show that individual average effort is highest for players in a tie position, followed by leaders, and lowest for followers. Starting as a follower (leader) leads to a lower (higher) chance of winning the race. Spillovers realized in the previous round significantly increase players’ investment in the current round. Convergence toward equilibrium play becomes more pronounced in the second half of the sessions. Efficiency loss is significantly higher in races starting from a symmetric position than from an asymmetric position and is also significantly higher in the low treatment than in the high treatment.