A system and method for optimizing a fuel efficiency of a hybrid vehicle by learning from a route history. The system may be a hybrid vehicle including an engine, a motor-generator, a battery, a battery module, a GPS unit, a memory, and a processor. The method may include detecting a current route of the hybrid vehicle, calculating a confidence value corresponding to a probability that the current route has a match in the route history, detecting a battery state of charge (SOC), determining a target SOC based on the confidence value and the route history, and adjusting an engine start/stop threshold and/or an engine power request to achieve the target SOC. They system may learn and improve fuel efficiency over successive trips along the route.