Starting with the simple Long-Short equities strategy as per Quantopian Lecture 39 “Long-Short Equity” template example, which combined Momentum, Quality & Value factors as the basis for ranking stocks, the following improvement was made, after adding constraints appropriate for the Quantopian competition (i.e. adding slippage & transaction costs, setting leverage to 1.0, etc.)
High absolute values of conventionally-calculated momentum constitute a necessary but not sufficient condition for good results with momentum strategies. What is required is not only a steep price trajectory but also a smooth one, and so adjustment of conventional momentum is required to incorporate this. Similar to the method described by Andreas Clenow in his book, part of this adjustment is made by incorporating an R-squared multiplier as calculated from the correlation between the input price series and a simple linear ramp. However, although this R^2 method of quantification of the general linearity or otherwise of the price trend is useful, it still does not incorporate any measure of the “spikiness” or smaller time scale “roughness” of the type of data specific to financial instrument time series. The author quantifies "roughness" using a simple but robust measure of the spikiness or bar-to-bar reversal tendency of the price data series. The originally calculated momentum (as per Lecture 39 template) is adjusted both for R-squared of price vs time, and also for the individual bar-to-bar spiky character (roughness) of the price data series. The resulting factor, denoted "Adjusted Momentum" is calculated as Momentum * R-squared * (1 – Roughness) and this is used in place of the original momentum factor. It is intended that this modification will lead to an improvement in the equity L-S momentum strategy and further development of this theme will be continued by the author.