A Pairs Trading Strategy for GOOG/GOOGL Using Machine Learning

We apply the spread model, the O-U model and SVM to build a pairs trading strategy for GOOG/GOOGL. There are two parts of the project that we think are novel after reviewing past related work: rst, we model not only price spread but also several selected technical indicators’ \spread"; second, we use two new metrics for measuring our trading strategy instead of the traditional back-testing method because we want to focus more on future prediction rather than return prediction, and to achieve that, we also propose two algorithms to reconstruct the data set. We outline the process of how our strategy is executed and, at the end, show that our strategy delievers a good win-rate.