Why are some soft-tossing lefties successful and others not?

When evaluating pitching prospects, fastball velocity is usually the best starting point. However, the minors are littered with soft tossing left-handed pitchers, a lot of them with good statistics. The Majors also have several left-handed pitchers with below average velocity that are enjoying quite a bit of success at the Major League level. I wondered if, by using Pitch F/X data, we can find any clues as to why some of these pitchers fail, while some of them succeed in the Majors.

So I looked at left-handed pitchers from 2007-2013 with at least 150 starter innings and a fastball between 87-89 MPH according to MLBAM tags, thus getting below average velocity pitchers, but not dealing with extremes like Mark Buerhle.

For a cut-off for which pitchers were successful, I considered ver 110 FIP not successful, under is successful. Obviously 110 is 10 percent below league average, and some comparisons of what that kind of pitcher that is statistically, Bud Norris, Ivan Nova, Phil Hughes, and Derek Holland in 2012 were near that mark. Basically guys that can fit into a rotation, but aren't what you would call excellent pitchers or even average. Jarrod Washburn and Nate Robertson were right at it, and I considered them successful.

In order to cut down on the brute amount of data to analyze, I pulled the start (if it was a Pitch F/X start) that was closest to each pitcher's career FIP, so for example, Rich Hill has a career 4.47 FIP, I pulled his 6/12/07 start, which had a 4.46 FIP.

It should be noted that none of them are even close to what you would call aces (there may not be a number 2 in the group either), and just a few of them were above average. However, the successful ones are pitchers that have some kind of value.

Successful lefties:
Dallas Braden
Ted Lilly
Tommy Milone
Chris Capuano
Jarrod Washburn
Chris Narveson
Eric Stults
Zach Duke
Nate Robertson
Randy Wolf
Odalis Perez
Paul Maholm
Andy Pettitte
Rich Hill

1253 overall pitches

Let's start with the basics:

I figured it had something to do with release point:
Vertical: 6.19
Horizontal: 1.98

 What about location?

628 pitches on the negative (glove side) part of the plate
625 pitches on the positive (arm side)part of the plate

Splitting the plate in half in height:
740 pitches below 2.5 feet
513 pitches above 2.5 feet

Unsuccessful lefties:
Scott Olsen
Ryan Rowland-Smith
Dontrelle Willis
Brian Tallet
Jeremy Sowers
Chuck James
Garrett Olsen
Jason Vargas
Greg Smith
John Lannan
Mike Hampton
Brian Burres

973 pitches

Vertical Release Point: 6.2
Horizontal Release Point: 2.00

498 pitches on the positive (arm side) side of the plate
475 pitches on the negative (glove side) side of the plate

431 pitches at 2.5 feet or above
542 pitches at 2.5 feet or below

So really, we see no difference in release point or general location between the "successful" and "unsuccessful" pitchers. That is, there are no easy answers, at least they don't lie in what I thought would be obvious differentiators. 

To look at pitches, we unfortunately have to use MLBAM tags here. So be sure to keep in mind that there has to be a decent margin for error here. When I think of lefties with below average fastballs, I usually think about curveballs or changeups, so I looked at those pitches for each group:

70 Unsuccessful pitchers curves, 7.2 % of all pitches.
Average spin: 306.92
Velocity: 75.1
Vertical Movement: -5.1

164 Successful pitcher curves, 13.1 % of all pitches.
Average spin: 308.03
Velocity: 72.52
Vertical Movement: -5.93

So we see that the "successful" lefties use their curves more, and while it has basically the same spin, their curves have a little bit better vertical depth and actually worse velocity.

229 Unsuccessful pitcher changeups, 23.5 % of all pitches
Vertical Movement: 5.87
MPH: 81.56

175 Successful pitcher changeups, 13.97 % of all pitches
Vertical Movement: 5.46
MPH: 81.07

So the unsuccessful pitchers actually have better movement on their changeups, along with slightly better velocity. However, they also use it a lot more, which makes me wonder if they overuse it. Out of those 229 changeups, 32 of them were for whiffs, well above average in swinging strike percentage. Compare this to the just 15 whiffs the successful pitchers got on their changeups. Weirdly (perhaps just because of the selective nature of the sample), the unsuccessful pitchers had a little bit better whiff percentage (7.5% versus 6.14%, both well below average). When you take out the changeups, the successful lefties have a very slightly better whiff percentage (5.51 % versus 5.75%).

Perhaps the changeup is what brought the unsuccessful pitchers to the majors, and as we have seen, when pitchers are known for having elite changeups coming up, that doesn't mean that the pitchers will have any kind of MLB success. While the pitch can be important to neutralize righties (for lefty pitcher), the fastball and curveball is more important. This is why the successful pitchers throwing their curves more and having better movement on the pitch seems meaningful (though it is strange that they are softer). Unsuccessful lefties got 6 whiffs with their curves, while the successful lefties got 13 whiffs with their curves in our sample.

I wanted to get a deeper look at location by looking at pitches down the middle of the plate, so I looked at pitches that were between 2 and 3 feet vertically, and .5 and -.5 horizontally. The successful lefties had about 270 of those pitches, while the unsuccessful righties had about 175 of these pitches. So strangely, the pitchers with the higher percentage of pitches down the middle were more successful. Of course, we have seen previously that the pitchers that throw the most pitches down the middle, have the most success down the middle, one of the seemingly many contradictions of pitching.

Other than the curveball/changeup distinction, using the data I used, I can't find a clear difference between soft-tossing lefties that have success in the Majors in starting roles, and ones that do not have success in starting roles. There are some things that are somewhat hard to measure using the data we have, such as sequencing, overall deception, and (because of pitch classification issues) fastball movement.

The best we can do with sequencing (trying to keep hitters off balance by using different pitches and fooling them) is look at which pitches they used with 2 strikes. With 2 strikes, unsuccessful lefties threw 33 sliders, some kind of fastball 126 times, 14 curveballs, and 46 changeups. Successful lefties threw 51 changeups, 49 curveballs, some kind of fastballs 173 times, and 62 sliders. Again, the big difference is the changeups versus curveballs.