I construct a series of time-series features from the literature and apply a novel XGBoost model to predict the next days price of a number of assets. The concept is simple and can be expanded to many variables, incorporate many assets and be applied to different Machine Learning models.