The accuracy of blast prediction with simple methods like Kingery and Bulmash’s curve-fits is limited by a number of errors:• The blast prediction method will not be perfect. When the prediction method is a curve that fits the average of measurement points, it is the scatter in the measurements themselves that determines the accuracy of the method.• The set-up of the real explosion, for which the blast has to be predicted, will differ from the set-up for which the prediction method is derived. For instance, there are likely to be differences in charge shape, explosive material, casing, weather conditions or ground surface conditions.• There are input parameters in the prediction method that are not always known. Guessing these will cause an error in the prediction.Knowing the accuracy of the blast prediction is important for e.g. scenario studies and reliability based design methods. This paper examines the contribution of each error source to the end result.The paper starts with a brief recapitulation of the definitions and the mathematics of accuracy. Next, the accuracy of Kingery and Bulmash’s curve-fits, the influence of unknown input parameters and differences between the real set-up and the set-up of the prediction method are discussed. It will be explained how all these different errors have to be added together to find the total error. It turns out that often the accuracy of blast prediction is limited by the error of the curve-fit (i.e. coefficient of variation, 15% for peak overpressure and impulse), but in some cases errors in the input can dominate (up to 100% error). The paper concludes by giving illustrative examples.