Music has previously been shown to be beneficial in improving runners performance in treadmill based experiments. This paper evaluates a generative music system, HEARTBEATS, designed to create biosignal synchronous music in real-time according to an individual athlete's heart-rate or cadence (steps per minute). The tempo, melody, and timbral features of the generated music are modulated according to biosensor input from each runner using a wearable Bluetooth sensor. We compare the relative performance of athletes listening to heart-rate and cadence synchronous music, across a randomized trial (N=57) on a trail course with 76ft of elevation. Participants were instructed to continue until perceived effort went beyond an 18 using the Borg rating of perceived exertion scale. We found that cadence-synchronous music improved performance and decreased perceived effort in male runners, and improved performance but not perceived effort in female runners, in comparison to heart-rate synchronous music. This work has implications for the future design and implementation of novel portable music systems and in music-assisted coaching.