How does the app work?
To find when future stock predictive power exists, we've explored the huge space of news sentiment interpretation, news half-life, stock price movement, and learning algorithms (i.e., AI).
With the right recipe we train the algorithm to historical stock price movement and news sentiment.
We then optimize the algorithm and half-life parameters using performance heat maps.
With optimal parameters selected, the app balances the portfolio of stocks using the precision of predictions in the unseen backtest period (i.e., validation data)
This provides the user with two backtest portfolio returns: even and optimal stock weighting. No algorithm is perfect, and we share when the algorithm performs well or not to better inform you.