Are we counting number of commits, or lines of code? If the latter, the calculation of the coefficient ought to take into account that larger commits almost certainly have more revisions (I'm assuming -- I haven't checked). efried .