Optimizing the Sharpe Ratio with Portfolio Turnover, Reasoning and Methodology

Abstract

This paper seeks to determine ways that a trading algorithm’s efficiency (specifically Sharpe Ratio over a given time period can be maximized through adjusting portfolio turnover in two ways: explicitly constraining the amount of turnover allowed per rebalance and rebalancing frequency. We test the first way via a Constraint method on Quantopian called “MaxTurnover(),” which, in lieu of sufficient explanation or documentation anywhere, seems to simply reduce average Turnover over a given trading period all else equal, of course. The second method of changing portfolio turnover is tested by simply increasing rebalancing frequency in the absence of “MaxTurnover()” as a constraint, id est turnover per rebalance is not constrained. From there, we examine ways we can generalize the data found. Current thinking dictates that extremely low turnover can be somewhat equated to missing out on too many opportunities and lowered exposure to alpha, but simultaneously that incredibly high turnover leads to too much exposure to risk. We find that accepted rules of thumb are still held true and that the data is consistent with current thinking and literature, but additionally that an unexpected relationship between drawdown and both measures of turnover--a positive correlation, that is. Equally unexpected is that this relationship is seemingly stronger than that between turnover measures and the Sharpe ratio, a proxy for portfolio efficiency. Through all of it, one point is clear, though: a modest rebalancing period and lower turnover is ideally efficient.