The paper presents a study regarding the selection of the most appropriate predictive method in assessing the athletes’ performance using the comparison of the residual value, referred to as unexplained variations. If the method used is perfect, then the unexplained variation is null. Alternatively, the average absolute deviation associated with each forecasting method can be used. In predictive analysis the future evolution of phenomena is estimated using different methods. The Markov chain method is a predictive model with limited utility that does not imply either the existence of a chronological series or the existence of an association. The Markov hypothesis assumes that the future state depends only on the present state, but also on a matrix of the probabilities of changing the state (the future state does not depend on past states) - the future is conditionally independent of the past. Another approach is the dynamic series or the time series that involves the use of historical data (records of the evolution of a phenomenon in time). The modified percentage method aims to evaluate the percentage change of the variable between successive periods of time. The mobile modifiedpercentage method has a higher degree of accuracy than the modified percentage method and it is used when trends are observed in the data. The method involves the prior calculation of the indices expressing the percentage change of the variable from one period to another. It also assumes the pre-calculation of the moving average of the modified percentage. The moving average method is used when we want to grant a higher important to recent observation of a set of historical data, compared to the beginning of the set.The method is based on the property of the average regarding the error compensation, thus diminishing the influence of periodic oscillations. The data represents the trend and reflects the common, general trend of the chronological series. The exponential smoothing method is more accurate than previous methods. In turn, it creates the possibility that the latest observations to be considered with higher rates. The method involves the selection of a smoothing coefficient whose value is determined either by using the moving average, or by testing and assessing the accuracy of the predicted values (sum of squared residuals).