Did I do my Sharpe ratio wrong?

[SoCalOptionsWriter]

1.18%
0.89%
0.56%
-1.53%
3.62%
0.43%
1.83%
2.14%
1.24%


Above are my monthly returns through September. I calculated an annual return so far of 10.64% (after additions and withdrawals in the account). I used 3.1 as a risk-free rate of return through 10.1 (three-quarters, about, of the one-year Treasury) and then divided by the SD (using the SD function in Excel) and came up with a Sharpe of 9.78, which seems ridiculously high. Process wrong? Math wrong? Thoughts?

 

[Sekiyo]

The math isn’t complicated,
If you respect the priorities (a-b)/c

The inputs look fine to me.
Even if you annualized 9 months …

I think 9 data points can yield ridiculous results.

 

[Gazillionaire]

You could try Sortino ratio...
https://www.educba.com/sortino-ratio/

 

[Businessman]

Ignoring the risk free rate for now, your annualised Sharpe extrapolated from those months is about 2.85

Average Month: 1.15
Stdev: 1.4

Annualised Sharpe = (Average/Stdev)*SQRT(12) = 2.85

[SQRT(12) is used because of 12 months in a year]

Your sample size is very small, so the long term Sharpe could in reality be lower than that.

If you factor in a 3% risk free rate: take 0.25% off every month, then i think the annualised Sharpe comes in around 2.23 using the same formula.

 

[syswizard]

Multicharts has the Sharpe Ratio as:
SharpeRatio = (AvgReturnPercent- int_rate_per_period) / ReturnStdDev;
There is no SQRT(12) component.

 

[Businessman]

There is no SQRT(12) component.

If the OP wants to calculate his yearly/annualised sharpe ratio from his monthly returns, then *SQRT(12) seems to be the way to do it.

If he wants the monthly sharpe ratio and not a yearly one, then he doesn't need the SQRT(12), but i am not sure how common it is to think in monthly sharpe ratios. Is it even a common thing?

Based on the above:

His monthly Sharpe is less than 1.
His yearly Sharpe is over 2.

 

[nrstrader]

I was going to say I think the period is .8 something if the risk free rate is zero.
For a risk free rate of 3.1 you would just subtract .258 from each return before averaging.

.8 sounds right to me as if those represented yearly returns it wouldn't be all that.

Of course, if this system ran for 5 years it would be quite nice but I think annualizing this is not giving a realistic standard dev when you don't even have a year of samples for an annual sharpe .

Getting 9 is using the sum returns divided by the monthly std dev instead of the average of the returns.

 

[Businessman]

.8 sounds right to me as if those represented yearly returns it wouldn't be all that.

If he can maintain that track record, 89% winning months, and also if the 11% losing months are relatively small, then his sharpe ratio is similar to Rentec Medallion. Medallion has a yearly Sharpe ratio of about 2.5. And around 90% winning months as well.

If the sample is legit representative of what he can expect, then the OP could just leverage returns 5x to make 60%+ a year and he would have returns very similar to Medallion.

 

[syswizard]

but i am not sure how common it is to think in monthly sharpe ratios. Is it even a common thing?
Based on the above:
His monthly Sharpe is less than 1.
His yearly Sharpe is over 2.

I don't see why monthly Sharpe would not be appropriate for high frequency trading systems ?

 

[nrstrader]

I don't disagree but that is also exactly why annualizing is not painting a realistic picture with 9 months of samples.
On a yearly basis it is trying to draw statistical conclusions from 1 sample.

If he can maintain that track record, 89% winning months, and also if the 11% losing months are relatively small, then his sharpe ratio is similar to Rentec Medallion. Medallion has a yearly Sharpe ratio of about 2.5. And around 90% winning months as well.

If the sample is legit representative of what he can expect, then the OP could just leverage returns 5x to make 60%+ a year and he would have returns very similar to Medallion.