Performance Evaluation of Systematic Option Trading Strategies Using Entry and Exit Points Predicted by Machine Learning

Abstract

In option trading, there are various technical indicators that can be used to determine entry signals. There are also various setups for systematic option trading strategies. The objective of this research is to evaluate the performance of systematic option trading strategies using entry signals generated by machine learning with technical indicators as predictors. To achieve this objective, a two-phase study is conducted. During Phase 1, machine learning algorithm is utilized to generate entry signals based on support and resistance prediction. Based on these signals, various option trading setups are evaluated in the Phase 2 to find an optimal setup that yield the highest ROC. Neural network algorithm with 5 hidden layers and 100 neurons in each layer is found to be the best algorithm to predict support and resistance to be used as signals. The back-testing results show that option trading strategy that yield the highest ROC is debit spread strategy with delta 0.15 setup and DTE 30 setup, which results in 399% annual ROC. The benchmark test results show that the performance of option trading using machine learning entry signals prediction outperforms option trading with single technical indicator entry signals, and that option trading outperforms the regular stock trading.