This letter introduces a fast procedure for predicting the performance of mechanical stirrers used in reverberation chambers (RCs) for over-the-air (OTA) tests of wireless communication devices, measured by the auto-correlation function (ACF) between stirrer-generated samples. The procedure is tested for three different stirrers in an RC where eleven load cases create widely different power-delay profiles, with a time spread spanning more than two orders of magnitude. The proposed procedure is shown to accurately predict the ACF and, therefore, how the number of independent samples is affected for each of the eleven RC load cases over an 8 GHz band, by only using data measured in unloaded conditions at 12 frequencies up to 4 GHz. Adding a second dataset with the RC loaded further reduces the uncertainty of the predictions, which are found to be more accurate than direct measurements, thanks to the regularizing effect of a wide-band parametric stirrer efficiency model. This method significantly reduces preliminary test time, providing a quantitative tool to evaluate beforehand the limitations of a stirrer's performance in OTA tests.