Implementing Trading Strategies for Forecasting Models ∗

In this paper we implement trading strategies for asset price forecasting models using parameterised decision rules. We develop a synthetic trading environment to investigate the relative effects, in terms of profitability, of modifying the forecasting model and the decision rule. We show that implementation of the trading rule can be as important to trading performance as the predictive ability of the forecasting model. We apply these techniques to an example of a forecasting model generated from an intra-day " statistical mispricing " of a combination of equity indices. Results indicate that optimisation of decision rules can significantly improve trading performance, with annualised Sharpe Ratio increasing by a up to a factor of two over a naïve trading rule. To achieve this level of performance increase through the forecasting model alone would require a 50% improvement in prediction accuracy.