Are Harder Curves or Slower Curves Better?

Mainly as a point of curiosity, I wanted to see if slow curves or hard curves worked better in the Majors. I expected to find, if there is any correlation between velocity and effectiveness in curves, that the harder and the slower curves prefered the best, while the ones closest to the mean were probably the worst. 

Brooks Baseball/Baseball Prospectus counts 254 pitchers who have thrown at least 200 curveballs since 2007. I looked at whiff/swing and GB/FB and split the curveballs down into different sections. The top 24 of the curves averaged 80 MPH or harder. Those curves averaged a 31.53 Whiff/Swing % and had a GB/FB of 2.61.
 
78-80 MPH: 28.29 Whiff/Swing %, 2.36 GB/FB
76-77.99 MPH: 25.98 Whiff/Swing %, 2.51 GB/FB
74-75.99 MPH: 24.51 Whiff/Swing %, 1.94 GB/FB
72-73.99 MPH: 25.35 Whiff/Swing %, 2.29 GB/FB
70-71.99 MPH: 24.14 Whiff/Swing %, 1.68 GB/FB
Under 70 MPH: 26.08 Whiff/Swing %, 1.75 GB/FB
So, at least by GB/FB, there is a pretty clear regression (other than the blip of 72-73.99 MPH) as the curves get slower. It isn't a leap to say that the harder the curve is, the more likely it is to cause ground-balls. When it comes to whiffs, the data seemed to somewhat support my original guess, the harder curves were the best, while the slowest curves were better than ones closest to the mean. However, it is pretty clear that the harder the curve, the more whiffs it got. 

Fangraphs allows us to easily break down pitchers into 30 pitcher chunks. I just looked at 2012. I looked wCU and wCU/C, and then went to each player's page to see their career wRC + against their curveballs were. I used a sample size of at least 80 for the wRC +.

Hardest 30: 82-86.2 MPH range, 13.4 wCU (.45 wCU per pitcher), 87.01 wCU/C (2.9 wCU/C per pitcher), 36.75 wRC +

Next 30: 81-82 MPH, 44.5 wCU (1.48 per pitcher), 27.37 wCU/C (.91 wCU/C per pitcher), 68.83 wRC +

Next 30: 79.9-81 MPH, 15.4 wCU (.51 per pitcher), -16.45 wCU/C (-.55 wCU/C), 64.8 wRC +

Next 30: 79.1-79.8 MPH, -12.3 wCU (-.41 per pitcher), -10.31 wCU/C (-.34 wCU/C per pitcher), 62 wRC +

Next 30: 78.5-79.1 MPH, 4.7 wCU (.16 per pitcher), -76.25 wCU/C (-2.54 wCU/C) 85.29 wRC +

Next 30: 77.9-78.5 MPH, -17.4 wCU (-.58 per pitcher), -20.81 wCU/C (-.69 wCU/C per pitcher) 68.44 wRC +

Next 30: 77.3-77.9 MPH, -1.3 wCU (-.04 per pitcher), -18.01 wCU/C (-.6 wCU/C) 73.04 wRC +

Next 30: 76.6-77.2 MPH, 4.97 wCU (.17 wCU), -13.92 wCU/C (-.464 wCU/C), 70.13 wRC +

Next 30: 75.8-76.6 MPH, -10.4 wCU (-.35 wCU), -22.02 (-.734 wCU/C), 71.43 wRC +

Next 30: 75.2-75.6 MPH,  -14.9 wCU (-.5 wCU per pitcher), -18.87 wCU/C (-.629 wCU/C), 70.41 wRC +

Next 30: 74.1-75.1 MPH, 27.2 wCU (.91 wCU per pitcher), 4.7 wCU/C (.16 wCU/C per pitcher), 67.68 wRC +

Next 30: 72.7-74.1 MPH, -6.6 wCU (-.22 wCU per pitcher), -25.34 wCU/C (-.84 wCU/C), 75.67 wRC +

Next 30: 70.6-72.7 MPH, 6.4 wCU (.21 wCU per pitcher), -40.22 wCU/C (-1.34 wCU/C), 67.89 wRC +

Last/Slowest 21: 63.2-70.6 MPH, -13.2 wCU (-.63 wCU per pitcher), -11.17 wCU/C (-.53 wCU/C), 72.05 wRC +

By wCU:
81-82 MPH
74.1-75.1 
79.9-81
82-86.2
70.6-72.7
76.6-77.2
78.5-79.1
77.3-77.9
72.7-74.1
75.8-76.6
79.1-79.9
75.2-75.6
77.9-78.5
63.2-70.6

By wCU/C:
82-86.2
81-82
74.1-75.1

79.1-79.8
76.6-77.2
63.2-70.6
79.9-81
77.3-77.9
75.2-75.6
77.9-78.5
75.8-76.5
72.7-74.1
70.6-72.7
78.5-79.1

By wRC +:
82-86.2
79.1-79.8
79.9-81
74.1-75.1
70.6-72.7
77.9-78.5
81-82
76.6-77.2
75.2-75.8
75.8-76.6
63.2-70.6
77.3-77.9
72.7-74.1
78.5-79.1

The correlation here is not as strong, but it does seem like the harder curves fared better than the rest. Overall, this data somewhat surprised me. I didn't think having a "hard" curve was a plus, in fact, I have been long been fascinated by the slow curves of Jered Weaver, Yu Darvish, and many others you see on the other side of the Pacific. This data suggests that it is better to have a hard curve, like King Felix, Matt Moore, or Stephen Strasburg. As long as these aren't classification errors (such as sliders being called curves), this has some big implications for scouting.

I did want to look at knuckle curves too, but according to Fangraphs, in the Pitch F/X era (2007-2012) only A.J. Burnett, Vin Mazzero, and Nathan Adcock (and maybe Chad Gaudin once) have thrown knuckle curves. So obviously there isn't enough data to look at.