# Calculations of statistics

Can someone tell me if there is some documentation on how the statistics are calculated for the strategies?

I am interested mostly to the more complex ones like Sharpe Ratio.
What period is taken for the calculation (whole history of strategy or just some latest period so it is more comparable between the strategies?),
What is the period for rate of return used there (daily, weekly etc.).
How does C2 calculate the Risk Free rate?

Anyone knows?

MS

We use daily returns of the strategy vs. daily returns of the SP500.
We analyze the entire strategy dataset - i.e. from strategy inception to current date.
Risk free rates change over time. Currently we’re using 2.83 (which admittedly probably needs some updating).

P.S. I’m speaking specifically about Sharpe ratio.

Thanks Matthew,

Do other indicators like Calmar, Sortino ratio also use daily returns?
Have you thought to introduce something like Sharpe ratio for the last year/2/3… to compare strategies? The longer the strat works the more difficult it is to maintain very high Sharpe, for new strat generating a high Sharpe is quite easy.
Same would apply to Sortino ratio.

MS

Yes, calmar, sortino, treynor all use daily returns.

Regarding adding new variants of stats for different time periods: I’ll regard that as a good product-improvement suggestion.

We may be reaching the point of diminishing returns when it comes to adding more statistics, but I can definitely see why this particular enhancement could be helpful.

Thanks Matthew,

I have studied the statistics in more details and I have some more questions, maybe you could help with these as well:

1. Then calculating Sharpe/Sortino/Calmar Ratio, do you take into account days where the position was flat? I compared your calculations on your page, with calculations in my trading software for the same period, and yours seem to be significantly higher than mine. That’s why this question came to mymind, perhaps that causes this difference.

2. How do you cope with geomethrical position scaling in your statistics (especially drawdown in %)?
My algo i always risking the same % of current capital on the account. In that case comparing the latest drawdown in \$ and then dividing by starting capital level makes no sens. The only proper way is to divide is to current highest level of capital.

MS

Sorry I don’t fully understand your questions, but let take a shot at quick answers:

1. Since we only compare percentage changes in daily values (that is, we compare account marked-to-market total value versus SP500) we don’t care “when the position was flat.” The only thing that matters (in the sense that it is the only raw material for the calculations) is the aggregate marked-to-market equity curve. Obviously that incorporates in some sense whether you are flat or not.

2. Any answer here depends on the specific statistic you refer to. So for example, if you are asking about MAE-to-trade-PL, or MAE-to-account-equity stats, we look at the minute-by-minute maximum adverse excursions (MAE) of the aggregate “trade sequence” (i.e. from flat to flat; this implies that the “closed portion” of the trade is part of the calculation), and we find the greatest MAE dollar value. We compare this value to either the final trade PL, or to the account equity at the moment of MAE. This stat’s usefulness may vary depending on trading style or whatever.