Hi All, Seeing if anyone is able to help me double check my Sharpe ratio calculations. For an example I've taken the daily returns of the S&P 500 Index (Data downloaded from Yahoo finance) from 1st February to 29th April (The time period i'm looking at). For the risk free rate i'm using 0.16% which is the risk free rate IB use for their calculations. - Since i'm doing daily returns I then divide this by 365 to get the daily risk free rate? For the period the figures I get are; Average excess return:0.1034% Standard deviation:0.87%. Dividing the average excess return by the standard deviation I then have an answer of 0.119 Since I'm using daily returns I then multiple this figure by the square root of 252( annual trading days) to get the final Sharpe ratio of 1.87. Does this seem right? In most examples it says to just divide average excess return by the standard deviation and that's your answer. The problem I have with this is that the Sharpe looks excessively low (in particular as the S&P performed very well during this time period). Interactive Brokers are also giving the Sharpe ratio for this time period at around 2.10 (relatively close to my figure). Any help appreciated. CG

Your figures are correct. But you have an incorrect opinion on what constitutes "low". Bearing in mind that the long run Sharpe of the S&P 500 is about 0.20, then 1.87 is not "low". GAT

Many thanks, I agree 1.87 is quite a high Sharpe for the market, I wasn't very clear but I was referring to the 0.119 figure as being too low for this time period considering the strong gains in the S&P during this time. Thanks again, CG

252 from a quick online search http://ir.theice.com/~/media/Files/...lume-reporting-tools/trading-days-2015-v2.pdf

Dividing the 0.16% annual risk-free rate by 365 is not quite the right way. The correct conversion can be derived from here: daily = (1+annual)^(1/365)-1 = (1+0.16)^(1/365)-1 = 0.000406713 Since you use 252 trading days for excess returns, you may as well use the corresponding risk-free rate adjusted for 252 days: daily = (1+annual)^(1/252)-1 = (1+0.16)^(1/252)-1 = 0.000589142 Your way is: daily = annual / 365 = 0.16 / 365 = 0.000438356 Since annual risk-free rate is so low, the resulting 3 answers above are not much different each other. In fact, for my purposes of calculating the Sharpe's ratio, I drop the risk-free rate from the equation altogether. Other than that, you calcs appear to be correct. Verify with this: Annualized Sharpe = sqrt(252) * [R / stdev(R)] = sqrt(252) * [(r-f) / stdev(r-f)] where r = daily asset (or index) return f = daily risk-free return If you drop the risk free rate, as I suggested, this simplifies to: Annualized Sharpe = sqrt(252) * (r / stdev(r))

252 is closer to being correct. 52 weeks*5 days in a week is 260. then subtract holidays. The NYSE lists 8 or 9 holidays. It depends on what day Christmas and New years land. 252 is correct.

To add to this topic, what is the norm for risk free calculation? I read this http://www.investopedia.com/terms/s/sharperatio.asp and they mention t-bill rates as a reference for risk free rate. My confusion is that wouldn't this number change all the time? What is the norm for using this for backtesting a portfolio strategy? I use Amibroker and it has the option to set the risk free rate which is by default 5.

In his much referenced paper, our friend William Sharpe does not actually use a "risk-free" rate in his formula, but rather the "return on the benchmark portfolio", such as the S&P 500 index: http://web.stanford.edu/~wfsharpe/art/sr/sr.htm It's not clear to me how and when the industry decided to use the T-bill returns instead of the stock benchmark returns. I'd speculate this was done to artificially inflate their Sharpe's ratios. Since a large number of funds under-perform the S&P 500 index, their Sharpe's ratios would all be negative, if calculated as originally specified by William Sharpe!

I use 256 because its 16*16 It makes no difference as long as you are consistent. This isn't a case where being strict makes any meaningful difference. It would change from country to country, and year to year, in any case. GAT