Relating reliability to analytical goals in performance analysis

The level of reliability of performance indicators used in an investigation should be related to the analytical goals of the investigation. This paper uses the example of analysing athletes using different tactical performances within the 2004 Olympic women's 800m running final. The percentage error statistic should not use the whole of the mean performance but consider a meaningful sub-range of values. There are different split times within the 800m that characterise the difference between the performances of different types of athlete such as leaders, waiters, breakers and pacers. A comparison of inter-operator errors with inter-sample differences for these split times is a useful means of relating reliability to analytical goals. The main contribution of this paper is the proposal of an alternative %error equation that considers absolute error as a percentage of the mean value recorded above some theoretical minimum value.