A picture is worth a 1000 words, right? This picture plots over 6300 pitches thrown by Carlos Marmol since 2007. Notice the drop in velocity followed by an increase in his slider velocity only (lower clusters).
While his fastball velocity has dropped, his slider has picked up speed and is back to as-fast-as-ever. One key for Marmol is having separation between his two pitches. Speed is just one aspect, movement is another.
PITCHf/x gives us both vertical spin deflection (or movement, when gravity is added back in as shown below) and lateral spin deflection (or movement). Guess what? Convergence.
This can't be good when combined with historically bad command. It's at least unsettling, especially when guys have teed off and taken Marmol deep at relatively alarming rate in 2011.
While Marmol's whiff rate on his slider is not horrible and actually "normal" (from 2008-2011: .34 .34 .47 .34) it isn't being put on the ground (GB rates: .48 .46 .46 .39) and isn't yielding pop-ups, his most under-rated weapon (IFF rates: .13 .12 .10 .06). As far as gopher balls, he's given up some on his slider for the first time since 2008.
Fastballs are a similar situation for Carlos. Whiff rates have returned from 2010's fantasy land (.23 .21 .31 .21) and remain exceptional for a heater. Balls in play have been a mixed bag and erratic against Marmol's fastball, except pop-ups which have evaporated this year (.26 .19 .16 .04).
Put it all together and you end up with a pitcher with a quality whiff rate (.31) which may not be as amazing as 2010 but still very good. Even his ground ball rate ends up being a wash, but there are has been a doubling in home runs per FB+LD and a halving of pop-ups per ball in play.
Take away those automatic outs in the infield, combined with a return from otherworldly whiff rates, and Marmol has fallen off the elite closer stack. Regain the separation, stop getting squared-up? I hope we get to find out sooner than later.