Ischemic stroke is a leading cause of disability and death worldwide among adults. Despite advances in treatment, around one-third of surviving patients still live with long-term disability. The individual prognosis after stroke is extremely dependent on treatment decisions physicians take during the acute phase. In the last five years several scores such as the ASTRAL, DRAGON and THRIVE scores have been proposed as tools to help physicians predict the patient functional outcome after three months of the initial stroke. These scores are rule based classifiers that use features available when the patient is admitted to the emergency room, and are selected or preselected by domain experts. In this paper we apply machine learning techniques to the problem of predicting the functional outcome of ischemic stroke patients, three months after they were admitted for emergency treatment. We show that a pure machine learning approach achieves only a marginally superior Area Under the ROC Curve (AUC) (0.798) than that of the best score (0.782) when using features only available at admission. Furthermore, we show that by progressively adding features available at other points in time, we can significantly increase the AUC to above 0.90, a value well above the scores. We conclude that the results obtained validate the use of the scores at the time of admission, but also point to the importance of using more features when possible, which require more advanced methods. We also conclude that machine learning techniques are instrumental tools that can be effectively used by physicians to improve the outcome prediction of ischemic stroke.