Eight different cylinder pressure trace based knock detection methods are compared using two reference cycles of different time-frequency content, reflecting single blast and developing blast, and a test population of 300 knocking cycles. It is shown that the choice of the pass window used for the pressure data has no significant effect on the results of the different methods, except for the KI20. In contrast to other authors, no sudden step in the knock characteristics is expected; first, because the data investigated contain only knocking cycles, and second, because a smooth transition between normal combustion and knock is expected, according to recent knock theory. It is not only the correlation coefficient, but also the Kendall coefficient of concordance, that is used to investigate the differences between the knock classification methods. It is found that none of the methods investigated possesses a significant advantage over the maximum amplitude of the bandpass filtered pressure method. Moreover, the methods based on maximum amplitude in the time domain agree reasonably well in their ranking, and the methods based on the average energy in the time domain produce similar results as well. No agreement between these two families is found.