Heart rate monitoring in physical exercise regimens is the key indicator of the workout intensity level. Day-to-day exercise variation of the heart rate reflects any progress achieved by the trainee and helps the trainer or the trainee himself to adjust the exercise work plan accordingly. However, timely decision upon changing intensity level of exercise is of crucial importance so as to maximize the health outcomes. Prediction of future heart rate values based on the trainee's history profile may prove to be a useful decision making tool in that case. The minimum set of available heart rate measurements in combination with the existence of outliers pose restrictions so to achieve reliable predictions. Time-series forecasting state-of-the-art algorithms such as Support Vector Regression and Gaussian Processes have been used in order to extract the best forecaster for these data. Heart rate data during and at the end of an exergaming intervention of 90 seniors were analyzed and compared in different cases. No single method outperformed the others. However, forecasting error was considered acceptable and all algorithms proved to be robust enough, even in the presence of outliers and irrespective the forecasting horizon, be it short or long term.