Hawk has a player/coach mindset. Advanced stats don't always make it easier to coach a player, you can get important marginal gains with BS like increasing TWTW (btw, if you don't think an increase in TWTW doesn't have something to do with Rios's improvement, you're wrong). There are also mechanical/technical adjustments, improved training, etc. that won't be quantifiable or especially predictable.
Sabermetrics are really good at explaining what has already happened. Sabermetrics can tell us that Adam Dunn was really valuable even though he struck out a lot and hit a lot of home runs. They may have told us he was trending towards more strikeouts. Unfortunately, Dunn's TWTW or something like that caused an amazingly precipitous downfall. Sabermetrics tell us that Gordon was probably unlucky last year and with neutral luck we should expect a somewhat better year this year. We'll see how accurate that is.
There is absolutely a tremendous balance to be achieved between quantitative considerations and qualitative. We are certainly getting better at quantifying things, though.