A system and method for personal fatigue risk management, performed using a smartwatch including a digital processor, associated memory, and an accelerometer for measuring acceleration in X, Y, and Z axes. A signal is received from the accelerometer indicating a movement of the smartwatch, and the value of the signal is stored in association with indicia of the time at which the signal was received in a time-stamped bin. The stored movement values are analyzed using a state algorithm to determine sleep, wake, and off-wrist state values. Fatigue risk values are determined from the sleep and wake state values, using a state algorithm. The probable relative risk of errors and incidents as compared to average risk, for the wearer of the watch, are then calculated. A warning is issued when the fatigue risk data indicates that one of the thresholds has been exceeded.