ABSTRACT
Chapter 9 considers “Market timing”, defined as the maximization of the signal-noise ratio of a strategy on a single asset, with conditioning information. Much of this material appears in more general form in Chapters 6 and 7. First the case of a single binary feature is considered. Under this model, the accuracy of the sign forecast controls the signal-noise ratio. It would appear that inference on the sign has higher power than a test of the Sharpe, but it is shown this is due to some questionable assumptions. Next, the case of market timing with a discrete feature is considered. Then it is shown that for market timing with a continuous feature the autocorrelation of returns and of the feature bound the Hotelling-Lawley trace of the strategy, which suggests that feature and returns must have similar autocorrelations. Finally a non-parametric trick for adjusting leverage is discussed.
