The aim of this paper is to introduce a two-step trading algorithm, named TI-SiSS. In the first step, using some technical analysis indicators and the two NLP-based metrics (namely Sentiment and Popularity) provided by FinScience and based on relevant news spread on social media, we construct a new index, named Trend Indicator. We exploit two well-known supervised machine learning methods for the newly introduced index: Extreme Gradient Boosting and Light Gradient Boosting Machine. The Trend Indicator, computed for each stock in our dataset, is able to distinguish three trend directions (upward/neutral/downward). Combining the Trend Indicator with other technical analysis indexes, we determine automated rules for buy/sell signals. We test our procedure on a dataset composed of 527 stocks belonging to American and European markets adequately discussed in the news.

We show how to compute the expectiles of the risk-neutral distribution from the prices of European call and put options. Empirical properties of these implicit expectiles are studied on a data-set of closing daily prices of FTSE MIB index options. We introduce the interexpectile difference , for , and suggest that it is a natural measure of the variability of the risk-neutral distribution. We investigate its theoretical and empirical properties and compare it with the VIX index computed by CBOE. We also discuss a theoretical comparison with implicit VaR and CVaR introduced in Barone Adesi [J. Risk Financ. Manage., 2016, 9].