According to Stalla, the definition of tracking error is: (sum(1-n) of [active return - mean active return] ^2 ) ______________________________________________ n-1 I dont get why the term “mean active return” is not “benchmark return”. What significance does “mean active return have”?

active return - mean active return squared gives you your sum of squared deviations, which you need to find the square root of to come up with tracking risk. CFAI has a couple nice examples, V4, pg 18 and pg 58 show the problem.

Nice. Thanks. I guess the issue was I forgot how to caclulate standard dev.