[sirana]

another answer, just to show, why a sample size of 50 hands is worth nothing if you are looking at Aggression factor (I'm figuring you take only the postflop aggression).
Assuming he's otherwise normal of the these 50 games he only played 35% = 17 hands.
Now lets assume that the half of these hands is rubbish on the flop, so he checks or folds them, which has no effect on the AF (aggression factor consists of (bet% + raise%)/call%)
So instead of 50 hands you are basing your play only on 9 hands.
If a majority of these 9 hands were drawing hands that missed, he would have an especially low AF, regardless of what his real playing tendencies are. If a majority of these hands were Top-Pair hands he will seem much more aggressive than he really is.
Furthermore, if he had only one single hand, where there was lots of raising from him and others his stats make him look like he was a complete maniac , even though he just correctly raised his quad, flush, straight...

just take it. AF is worth nearly nothing if your sample size is small and if your sample size is bigger you should have gotten some good reads all the time you played him.
pockertracker stats are just no substitute for real reads...