Glossary
Covariance
A measure of how two return series move together, in units of return squared. Correlation is covariance scaled to lie between -1 and 1.
The covariance of returns x and y is the average of (x - mean_x)(y - mean_y). It equals correlation x sigma_x x sigma_y. A positive value means the two tend to be above or below their averages together.
The variance of a portfolio is the weighted sum of all pairwise covariances, which is why diversification depends on how assets move together. A covariance matrix for many assets from a short sample is noisy.
Example: two assets with volatilities of 20% and 12% and correlation 0.3 have a covariance of 0.3 x 0.20 x 0.12 = 0.0072.