A Stochastic Asset Price Model with Adaptive and Asymmetric Mean Reversion

Authors:

Abstract
We propose a stochastic model for short-term asset prices movements which, alongside traditional dynamics, features mean reversion towards a calibrated trailing average. The model allows for a different strength of reversion dependent upon whether the current price is above or below the trailing average. We fit the model via Bayesian methods to a dataset of stock prices spanning more than one decade. We show unequivocally that the inclusion of calibrated mean reversion greatly improves the fit of the model, with the strongest effects observed at short time frames e.g. hourly data. We also demonstrate a simple trading strategy based on the model; we show that this strategy has returns comparable to that reported elsewhere via sophisticated machine learning approaches, with the advantage that it relies on only a single, easily interpreted feature of the model.

This publication has 38 references indexed in Scilit: