7
$\begingroup$

I have a pair trading strategy with positions that last 3-5 days and trades 2-3 times a month. By design, all the trades are profitable until the cointegration is broken.

Should I calculate the Sharpe ratio with daily or monthly returns? (annualizing afterwards in each case)

With monthly returns, most positions will be closed so I will have mostly profits (and maybe a loss if there's an open position at the end of the month).

With daily returns, I will have partial profit and losses each day.

I haven't done the calculations yet, but seems to me that the annualized Sharpe ratio of the monthly returns will be higher than the one with the daily returns, even with the difference of the annualization factors $\sqrt{12}$, $\sqrt{252}$ respectively.

$\endgroup$
6
$\begingroup$

Sharpe should only be computed from daily returns because finer granularity leads to a larger sample size. The larger sample makes the standard deviation metric more accurate. As a counter-example, how reliable would the Sharpe be using yearly returns?

$\endgroup$
  • $\begingroup$ It also has to be done that way because almost everyone out there is doing it on daily returns. This enables clearer comparisons between strategies. $\endgroup$ – Zarbouzou Jun 10 '12 at 9:14
  • $\begingroup$ @Zarbouzou that's incorrect: Yes everyone scales Sharpe ratio to annualized terms, but it can be scaled from any other sample duration (using "annualization factors" as noted in the original post). $\endgroup$ – feetwet Jul 14 '14 at 21:28
  • $\begingroup$ By this argument, shouldn't one use microsecond returns? $\endgroup$ – steveo'america Jun 20 '18 at 18:20
2
$\begingroup$

I think that it is better to look at t-stat than at the Sharpe Ration itself with data using different frequencies (thus different df) in order to determine which Sharpe Ratio stat is more accurate.

Calculation can be found in this paper: The Statistics of Sharpe Ratios.

$\endgroup$
  • $\begingroup$ Could you give an example? I didn't understand how to compare frequencies from the paper you linked to. $\endgroup$ – chrisaycock Jun 11 '12 at 0:38
  • $\begingroup$ Calculate SR. The SE is given in Eq (9). To get the t-stat SR / SE(SR). Note the T in the Eq. is sample size. I hope this helps. $\endgroup$ – Suminda Sirinath S. Dharmasena Jun 11 '12 at 7:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.