Can somebody confirm this? Let's say I have a hypothetical return of average 1% per month and that my monthly return standard deviation is also 1%. Then most of the monthly return is somewhere between 0 to 2%. The annualized Sharpe ratio would then be SQRT(12)*1/1 = 3.46. That's quite a high ratio given that my strategy only returns 12% per year. So is that right?
Your formula is correct, so probably your values plugged in are wrong. Returns are clear cut so your stdev is probably wrong. Easier way, take daily returns Calculate std dev of this data series Multiply by SQRT(256) = 16 That is your sharpe ratio.
That was a just a hypothetical average return and standard deviation. But I do understand now why HFT typically has higher Sharpe ratio. Just the sheer number of trades (even with just a little bit alpha) can achieve smooth monthly earnings.
Very nicely explained. In theory, the math for the numerator of the Sharpe ratio in the context of a market neutral strategy works out to: (Long - risk free) - (Short - risk free) = Long - Short, as risk free cancels out.
Sharpe ratio does not tell anything about the return on capital, but rather it’s an indicator of signal to noise. If you are uncertain about the numbers, go ahead and calculate sharpe ratio on dollar PNL, you should get similar value.
The formula that you cite is from the original formulation, published in 1966. Sharpe has revised it in 1994 to this: Sharpe ratio = (Average Portfolio Return - Risk free rate of return) / STDEV(excess returns) If the risk free rate is 0, the 1994 version reduces to the 1966 version. Yes, if you want to be precise, you should use a separate risk-free rate for each period. For a simplified method, assume risk free rate to be 0, which reduces the formula to: Sharpe ratio = Average Portfolio Return / STDEV of returns Then you need to annualize it, as suggested in the previous posts.
Finally getting around to adding Sharpe Ratio to my platform. One more question. Assume I had the following returns: 2015: 12% 2016: 18% 2017: 19% 2018: 23% 2019 (YTD): 4% How should I handle 2019 since the year has not completed and it introduces a misleading deviation in returns. Should I do something like (365/CurrentDayOfYear) * 4% to extrapolate the YTD return to a yearly value? thanks
What you should do is look at periodic returns (daily, weekly, monthly, whatever you think is applicable), calculate the Sharpe ratio for that return periodicity and then rescale it to the annualized form. E.g. if you have daily data Code: annualized_sharpe_ratio = sqrt(252) * mean(daily_returns)/stdev(daily_returns) PS. There is a lot of little things that surround S/R that are worth implementing too, like confidence intervals given the sample size, Sharpe decay rate etc