Whether driving a car or playing a sport, humans often make rapid decisions based on the actions of others. Decision-making has been studied in single participant tasks by manipulating the uncertainty of visual information used to make the decision. Yet, to our knowledge, using visual information of another person's actions to make rapid decisions remains unstudied. Here we test the idea that humans accumulate evidence of their opponent's intended action to make a rapid decision. Each participant, of a human pair, moved a robotic manipulandum to control a cursor from a start position to one of two targets. Both participants had vision of their opponent's cursor. The two players were assigned as either 'predator' or 'prey'. The predator and prey had competing goals of reaching to the same target and different target, respectively. We manipulated the time available to reach to a target in three conditions: 500ms, 850ms, and 1500ms. Hand paths early in a trial were less predictive of target selection, suggesting participants exploited more available time in the longer trials. Further, participants were more successful when they delayed their final change in cursor movement direction and relied on a rapid decision near the end of a trial. Taken together, these results suggest that the participants exploited additional time to accumulate evidence about their opponent's intended action, and relied on rapid decision-making when reaching to multiple potential targets in a competitive scenario. Preliminary work shows that a drift-diffusion model captures temporal decision-making of competing humans.