<p>It is essential to combine mathematical–statistical methods with the acoustic emission (AE) parameter to explore an effective precursor of rockburst. In this study, unloading rockburst experiments are performed and the evolution of AE counts is illustrated. Next, the mutation point of the AE counts is confirmed using the Mann–Kendall (M–K) test, and the influence of the window length and lag step length on the autocorrelation coefficient and variance are studied based on the critical slowing-down theory. Afterwards, the universality of the M–K test and critical slowing-down theory is analysed based on common experimental methods and rock types. Combined with the traditional indexes, the recognition ability of the precursor is evaluated. Finally, the rockburst ejection process is elaborated based on the critical slowing-down theory. The results reveal that the mutation point obtained by two different methods can predict rockbursts, and the sudden increase in variance can be considered as a robust precursor that indicates rockburst. Furthermore, the precursor derived from the critical slowing-down theory is earlier than those obtained from the b-value and information entropy, whereas the precursor from the M–K test is the latest, both serving as long- and short-term warning precursors, respectively.</p>