In most sources, the covariance of matrix of observations $X \in (n,p)$ is usually defined as
$$cov(X) = \frac{\hat X^T \hat X}{n-1}$$ where $\hat X = X - \mu (X)$.
However, after reading the definition of Wikipedia, it is not divided by the $(n-1)$ term. Why? I'm not talking about bias, since they don't divide by $n$ either.