On this page we present various interesting research articles with focus on performance measurements. Given the focus of the site, most of these will have some bearing on how to evaluate hedge fund strategies.
Sharpe Ratio and Critique
The illusion of skill (Wayne Himelsein and David Taylor)
The fallibility of Sharpe Ratio, and all of its cousins, becomes evident upon recognizing that it relies solely on the mean and standard deviation, which, as summary statistics,cannot distinguish between processes that are symmetric and asymmetric. In order for the Sharpe ratio, and other related statistical tools, to accurately measure a risk reward profile, the underlying time series has to be symmetric.Most users of Sharpe and related tools assume this is the case. However,the standard across the vast majority of investment strategies is quite the opposite of this assumption; returns are most generally asymmetrically distributed, and particularly, negatively skewed. Moreover, erroneously assuming that an asymmetric distribution is symmetric when it is left skewed,will cause any model to dramatically understate risk.This is especially true when underlying distributions exhibit high levels of negative skewness, which lo and behold,is most often the case.Download the paper here: https://www.logicafunds.com/blog-research
An interesting paper on the difficulties comparing different strategies using a common metric, i.e. the Sharpe Ratio. The problem here is that without knowing the higher moments of a distribution, it is difficult to use the “Skill Metric” that the authors have developed.
Several other papers have come up with variations of the Sharpe ratio and how to address the shortcomings of the measurement. In the end, the Sharpe Ratio is only a model and the model is only good as long as the assumptions hold. This paper broadens out our understanding of how to apply the tool in situations when the distribution of returns is not close to normally distributed.
Drawdowns (Otto van Hemert, Mark Ganz, Campbell R. Harvey, Sandy Rattray, Eva
Sanchez Martin, and Darrel Yawitch)
Common risk metrics reported in academia include volatility, skewness, and factor exposures. The maximum drawdown statistic is rarely calculated, perhaps because it is path dependent and estimated with greater uncertainty. In practice, however, asset managers and fiduciaries routinely use the drawdown statistic for fund allocation and redemption decisions. To help such decisions, we begin by quantifying the probability of hitting a certain drawdown level, given various return distribution properties. Next, we show that drawdown-based rules can be particularly useful for improving investment performance over time by detecting managers that lose their ability to outperform. This can happen as a result of structural market changes, increased competition for the type of strategy employed, staff turnover or a fund accumulating too many assets. Finally, we show that drawdown-based rules can be used as a risk reduction technique, but this impacts both expected returns and risk.Download the paper here: https://papers.ssrn.com/sol3/Papers.cfm?abstract_id=3583864
An in-depth study of the drawdown measure, how to turn it into useful rules, but that returns and risks are impacted. The implicit audience of the paper is a fund-of-fund manager that bases his/her decision upon predefined loss levels.
They also manage to establish that your largest drawdown is always ahead of you and that rules based on maximum drawdown should be time-sensitive rather than static. We may replicate their finding using our database.