1) I make the sample size small enough to get to the top 30.
2) I don't just look at UZR/150, and when I do, I value 3 years over one most recent.
3) I also look at DRS, which has Bogey dead last in any sample size you choose to use.
To me, put it all together, and the numbers say, "Below average." That happens to jive with my own personal observations, as well.
I also do not see the defensive "improvement" many speak of.