# Sharpe ratio from second returns? HFT

I have an intraday trading strategy so I take several positions each day. I have prices for each stock, with a 10 second resolution. So the data looks like this:

09:00:00   $$100 09:00:10$$101
09:00:20   $$99 09:00:30$$97
...


So I can calculate the return for each 10s interval. Then I sum the returns up, to get the total return for that day. My question is: how do I get annualized Sharpe ratio from these 10s returns?

Right now I am experimenting with one days worth of data. At the end of the day, I get a return of 183% (just experimenting with some random data and random trades). If I annualise my returns by multiplying with sqrt(252) I get a Sharpe of 0.76. That does not sound right? If I get a daily return of 183% then the annualized Sharpe should be much higher than 0.76? So what I am doing wrong? I am following this: How to calculate Sharpe Ratio from \$ returns?

• Roll them into daily returns and then compute the Sharpe on daily returns "in the usual way". There is really no gain in statistical accuracy from using returns at the second level. This is discussed in Chapter 3 of Short Sharpe Course on SSRN. Mar 26 at 23:45
• If you haven't done it, try and recompute everything using log-returns, since log-returns aggregate over time by summation. Also 2 things come to mind when calculating the Sharpe ratio: firstly, if you take an excessive amount of risk in your trading strategy then it makes sense that your Sharpe is "low", and lastly, if you did not subtract your transaction costs (or any costs whatsoever) from your returns, then you might get a higher Sharpe than intended.
– Pleb
Mar 27 at 9:37
• Thanx for your help! I have transformed the returns into logarithmic returns (as described in my link). So I can add up the returns for each day, and arrive at a daily total return. This daily return I can use as usual: calculate the mean and std dev, and multiply with square root of 252 - this gives me annualized Sharpe. However, I wonder out of curiousity, how would I transform my 10 sec returns, into an annual sharpe? I want to understand the theory better. I just calculate the average 10 sec returns and std dev, and multiply with the root of how many 10s periods there are in a year? Mar 28 at 15:43