Candlestick Analysis in Forecasting U.S. Stock Market: Are They Informative and Effective

Stock price prediction is one of the hottest topics of research in both academia and industry. Being able to predict the trend of price correctly allows investors to gain profit. There have been multiple strategies in stock price prediction, such as multiple machine learning methodologies and forecast from sentiment analysis of the public and news feedings. Among these strategies, one of the oldest but still widely used strategy is the candlestick analysis, which is a simple way that allows general investors to predict the market trend. However, there lacks an unbiased estimation of the effectiveness of such a method. In this paper, using an unbiased and rigorous way to test multiple U.S. stocks, we were able to show that most of the candlestick patterns were not informative, while a small fraction of them provides some correct information for market trend compared with random guesses. Our study serves as a stepping-stone for re-evaluating the candlestick analysis and urges more similar and thorough studies to be conducted to guide the general public in stock market investment better.