Optimal statistical arbitrage: a model specification analysis on ISEQ equity data

A comprehensive empirical analysis of the novel optimal statistical arbitrage trading model of Bertram (2010) is performed on a dataset of stocks quoted on the Irish Stock Exchange. Evidence of significant errors on average in the key measures underlying the trading model is presented, reflecting the mis-specification of the underlying Gaussian Ornstein–Uhlenbeck (OU) process. Overestimation of the expected return per unit time measure and underestimation of the expected trade cycle time measure are most notable. It is further shown that the Bertram (2010) trading model is more robust to high mean reversion and/or volatility parameter estimates compared to two benchmark models based on the exact and approximate first-hitting time densities of Linetsky (2004) for an OU process.