There is no doubt that sentiment affects decisions and thus market prices. Our trading system uses Twitter data to measure the sentiment for 26 stocks included in the S&P 500 stock market index. By analyzing the sentiment, the system is able to exploit behavioral biases.
The main feature of the system is the word analysis algorithm that derives a sentiment measure for each stock. Long positions are taken in the three stocks with the most positive measures and short positions are taken in the three stocks with the most negative measures.
By testing different formulas for calculating the sentiment measure, the formula as well as the other parameters affecting the trading decision have been optimized. The other parameters affecting the trading decision are the number of followers an account needs to have in order to be considered reliable and how many hours to consider before the market opens.
So far, the trading strategy has generated abnormally high returns. On average, the daily return is 0.38% while carrying a low risk. Jensen’s alpha is positive at a 1% significance level, implying that the strategy generates excess risk-adjusted returns. In addition, the beta is extraordinarily low (not statistically different from 0). In other words, the portfolio generated by the algorithm has no or little systematic risk. In almost 72% of the trades, the system is able to predict the correct sign.
Analysts: David Frykmer and Abdeljalil Moussa
