Correlation instead of Covariance is used in Calculating Portfolio Variance?

[chuhean]

In the function estimate_SR_given_weights() for static optimisation of the best set of instruments, there are 3 variables wt, mu, cm used to calculate the portfolio SR with neg_SR().

The formula used to calculate the variance requires the weight and covariance. However, the code seems to use weight (wt) and correlation (cm), instead of covariance.

[robcarver17]

You are correct; however the standard deviations that are fed in are all unit 1 (since these are substrategy returns, so they are equal standard deviation in expectation) so fixing this won't actually change behaviour. Still worth fixing though.