What are the typical metrics used to evaluate hedge fund performance, and what are their pros/cons with respect to each other? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world.
Here are my thoughts on some of the main metrics that are used to evaluate hedge fund performance:
Return--Return is meaningless as a standalone statistic. Want a simple proof? Assume a trader or money manager with net positive returns doubles her exposure on every trade. Her net return will double. Are the results suddenly twice as good? Of course not. All the loss statistics will also double. Return only has meaning relative to the risk level taken to earn that return. That is why I never look at return in isolation.
Sharpe ratio--The Sharpe ratio is the most widely used return/risk statistic. It is a good measure but does have flaws. The most significant drawback in the Sharpe ratio is that the risk component of this metric--standard deviation, a volatility measure--does not differentiate between upside and downside volatility. Thus, a sharp gain will adversely impact the risk component exactly as much as an equivalent large loss. The reality is that people's intuitive sense of risk is based on losses, not volatility. Investors will get very upset when their manager loses 15% in one month; I have yet to meet the investor who gets upset when their manager gains 15% in one month.
Sortino Ratio--Without getting into the math, the Sortino ratio is similar to the Sharpe ratio except that it penalizes only downside deviations, not all deviations. In this sense, it is an improvement over the Sharpe ratio.
As a tangential comment, in my opinion, the way most sources calculate the Sortino ratio is wrong. Essentially, they calculate the risk component assuming the number of data points is equal to all data points even though only data points with returns below the "minimum acceptable return" (typically assumed to be zero) are used. As a result, the Sortino ratio will almost invariably be higher than the Sharpe ratio even for managers whose returns are negatively skewed (that is, the largest losses tend to be of greater magnitude than the largest gains). At FundSeeder, we correct for this distortion by dividing the conventionally calculated Sortino ratio by the square root of 2. Readers interested in more detail can check the site.
Maximum Drawdown--The maximum drawdown is the largest percentage decline from an equity peak to a subsequent low. This is a good statistic to look at, in the sense you want to know the prior worst case for a manager. Of course, there is no guarantee that a future drawdown won't exceed the maximum drawdown. Since this measure only reflects one event in the data, I would consider it a supplemental statistic, never a primary one.
Calmar and MAR Ratios--These are return/risk measure that use the maximum drawdown to represent risk (instead of the standard deviation, as in the Sharpe ratio, or downside deviations, as does the Sortino ratio). I am not a fan of these metrics as primary return/risk measures because the risk component only considers what is typically a minor portion of the data. I would rather use other return/risk measures and then view the maximum drawdown alone as a supplementary statistic.
Gain Pain Ratio--This statistic is not typically used but I am including it in this discussion because it is my single favorite statistic. The GPR is the ratio of the sum of the returns to the absolute value of the sum of all the negative returns. It basically tells you how much return was made relative to all the losses incurred to get that return. For example, assume we are using monthly data, if a manager has an average annual return of 10% (arithmetic not compounded return) and a GPR of 1, it means, on average, the manager experiences a total of 10% in losses per year across all losing months to achieve that average net return of 10%. I like this statistic because it penalizes all losses, and only losses.
Ideally, the calculation would be done using daily data instead of monthly data. Daily data is much more informative, not only because it provides about 22 times as much data, but also because it will reveal losses hidden by monthly data. For example, if a manager losses 15% and then rebounds and finishes the month up 1%, the monthly GPR will be positively affected, whereas the daily data will reveal the risk hidden by the monthly data. For this reason, monthly GPRs will always be much higher than daily GPRs. Empirically, I have observed the ratio between the monthly and daily GPR averages about 7:1, although the ratios can vary widely among managers. Roughly speaking, a daily GPR above 0.15 is good and above, 0.30 is excellent. For monthly data, roughly speaking, a GPR above 1.0 is good and above 2.0 is excellent.
If you have read this far, you would probably interested in knowing that I recently completed two video presentations on the subject of performance measurement for the upcoming . These videos, as well as all the videos in the summit, will be available for free viewing on February 8 (or accessible anytime for a fee).
Also, all the performance statistics mentioned in this reply and many others, as well as a range of performance graphics and analytical tools are available for free by linking your account to .
This question originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world. You can follow Quora on Twitter, Facebook, and Google+. More questions: