Branch prediction is the process of predicting the outcome of conditional branches before they are actually executed. Branch prediction techniques are divided into two broad classes: static and dynamic. Static branch prediction associates a fixed prediction to each static branch at compile time whereas dynamic strategies makes a prediction at run time. Today`s most accurate static method is evidence-based static prediction (ESP), which uses machine learning to generate a prediction based on a set of static features. By extending the static feature set of ESP, we further reduce the miss rate of 17.4% for ESP towards 11.6% (SPEC 2000). Although almost all dynamic predictors do better than static methods, highly accurate static prediction can serve as prediction for weak dynamic predictions in e.g., gshare.