Improvements in the results of athletic competitions are often considered to stem from better training and equipment, but elements of chance are always present in athletics and these also contribute. Here we distinguish between these two effects by estimating the range into which athletic records would have fallen in the absence of systematic progress and then comparing this with actual performance results. We find that only 4 out of 22 disciplines have shown a systematic improvement, and that annual best results worldwide show saturation in some disciplines.