Stochastic methods use mathematical — above all statistical — tools to forecast values or to signal that significant events are approaching.
Regression models
The most widely used statistical tool is regression, linear and non-linear, which models current prices from previous prices and volumes. These models come with a very reliable measure, R-squared, which shows how well the model predicts yesterday’s prices from earlier days’ information. There is, of course, no guarantee for tomorrow’s prices, especially when sudden economic changes cause shocks — which is why it is always good practice to pair regression models with natural-language news analysers. More sophisticated models also exist, capturing several layers of interaction to better reflect economic reality.
Genetic algorithms and neural networks
Other stochastic methods include genetic algorithms, which predict prices with a self-adapting strategy inspired by biology, and neural networks, which use a network of simple nodes that builds its own economic model of prices.
