A Test of Market Efficiency: A Supervised Machine Learning Approach to Binary Options Trading

Abstract

Joe Wayne Byers, Ioannis Evgeniou and Anand Ravindra Manjrekar

There has been significant progress in automating binary options trading and in developing more sophisticated trading strategies. Most of these tools rely on historical data of shorter time frames; often without verifying the presence of any predictable patterns. These trading systems lack robust predictive analytics, which results in exposing the retail traders to a significant risk. This study attempts to address this loophole through a comprehensive analysis using large datasets to detect the presence of any predictable patterns. The study involves the usage of a sophisticated machine learning technique such as XGBoost. The workflow commences from generation of datasets to feature extraction, then it is followed by patterns recognition for determining trading positions. The features and outcomes are stored and used for the hyperparameter tuning process which is followed by the model training process. The trained models participate in a forward testing simulation for prediction of trading position entries. The key performance metrics of win rate and total number of trades are calculated and displayed. The results will aid in investigating the existence of any predictable patterns within shorter time frame data with the usage of a trading bot that is powered by XGBoost. The findings of this research could significantly impact regulatory policies by showing a need for much stricter regulations over the minimum time frame involved for binary options trading. It would underline the necessity to protect the retail trader’s community from potential fraudulent trading activities from algorithmic trading firms that claim predictive power. This work will arm retail traders with the best practices and provide sufficient basis for informed trading decisions.

PDF