Automating measures of performance can allow for heightened understanding of how an athlete is performing, not only during a session but over time. Physical and mental performance degrades in high-intensity skilled sports like boxing as the athlete fatigues. Although the use of low cost, ubiquitous inertial sensors have been reported effective for performance classification in boxing, no assessment of a boxer's efficacy has been reported under fatigue conditions. This letter evaluates the use of inertial sensors for automatic classification of fatigue by assessing the punch consistency in terms of pitch angle, punch force through acceleration, and hand speed, using the inverse time between punches. To achieve this, bespoke software was created in MATLAB, aided by the use of an attitude and heading reference system orientation filter. Six right-handed male elite boxers from the Pacific Nations, in preparation for the 2018 Gold Coast Commonwealth Games, Australia, consented to participate in the study. A noticeable decrease in performance for both hand speed (inverse time between punches) and force production (acceleration) was observed over time during intensive training, resulting in a Pearson's correlation coefficient of r 0.97 for the acceleration component and r 0.89 for the timing component. Inertial-magnetic sensors, with bespoke software, were experimentally found to be an effective tool for the automatic classification of boxing fatigue performance metrics. This study was conducted under ethical approval (ENG1413HREC).