Quality score for MC PageRank approximation

Hey people, I'm implementing a Monte Carlo PageRank approximation. I want to measure it's accuracy when compared to the "true" scores that power iteration gets. So I have two probability distribution vectors that I compare.

I'm trying to come up with a metric that should show what is expected for this method, that is that the values are close for the larger values in the vector (ir. more important PageRank values are approximated closer to their true value) and less so for the smaller PageRank values.

Any ideas?