If we know anything about preventing runs and hits, it's that the defense functions as a unit. The pitcher is responsible for a large portion of preventing damage simply by avoiding walks, homers, and hard contact, but there's a lot that goes into the mix that he can't control. The quality of the defense behind him matters. The ballpark he calls home matters. The simple distribution of batted ball luck matters.
Two pitchers can allow the exact same batted ball off the exact same pitch and find two different outcomes, is the point, and we've long known that. One pitcher may allow a hit and a run (or runs) and the other may get an out, but did they really do anything different? Was the skill different? Perhaps not.
In order to find out, we can use Statcast™ to get away from outcome-based stats like batting average and ERA, and instead get towards process-based stats. What we mean by that is, how much exit velocity did the pitcher allow on a given batted ball? At what angle was it hit? Most importantly, what's the expected outcome on batted balls with similar characteristics across the sport? By comparing a pitcher's expected production against his actual, we can see who overperformed or underperformed in 2016, and try to identify what factors may have caused that.
Depending on the reason, this could be a good way to find rebounds to count on in 2017… or unsustainable performances to worry about.
Here's how we did this, distilled down to three important points:
1) The metric of choice is Weighted On Base Average (wOBA), a stat that's similar to on-base percentage except that it weights extra-base hits more heavily than singles, rather than counting all times on base equally. The 2016 Major League average for wOBA on contact was .363; obviously, we're just looking at batted balls, so strikeouts and walks are excluded.
2) Each batted ball was assigned an expected wOBA based on velocity and angle, which accumulates into a player's seasonal average. Batted balls were also grouped into six categories, explained here, like "Barrels," "Weak Contact," "Sky High Balls," etc.
In a way, this helps account for park factors to some extent. For example, this Jose Altuvehome run off ofChris Archer (expected wOBA: .206) was worth a fraction of this Altuve blast off of A.J. Griffin (expected wOBA: 1.878), because the ball Archer allowed went only 340 feet and is an easy flyout in most parks that don't have short left field porches like Houston does.
3) The limited subset of batted balls that were not tracked by the Statcast™ radars were estimated using a process described by MLB.com's Tom Tango here.
Got that? We also looked only at starting pitchers who allowed 100 batted balls. (For reference, the starter who had the highest estimated wOBA, at .458, was Christopher Young; the lowest, at .307, was Steven Wright.) That in mind, let's check out our five "luckiest" and "unluckiest" pitchers of 2016 -- knowing full well that it could be a number of things other than "luck," like defense or horizontal angle -- and maybe learn a little about why.
2016's "LUCKIEST" STARTERS
5. (tie) Martin Perez, Rangers
Estimated wOBA: .365
Actual wOBA: .321
Difference: .044
The first thing to know about Perez is that he simply does not miss bats, and that's not hyperbole. His 12.1 percent strikeout rate was the lowest of any qualified pitcher in 2016, even lower than Jered Weaver's 13.4 percent. You might notice that his expected wOBA of .365 was just about exactly the same as the Major League average of .363, and a pitcher who allows average contact with below-average ability to get strikeouts would be expected to have below-average results, and Perez did, putting up a 4.39 ERA.
But "below average" isn't "terrible," and Perez was still decent. How? Perez's performance on low-impact batted balls wasn't all that different from expected, but on what we're calling "solid contact and barrels," his expected wOBA of .975 was only .781 in reality, a huge difference. There's a few reasons why that could be, and we can't say we know for sure yet. However, one interesting point is that in the span of three batters in Detroit on May 8, Perez allowed a pair of 100 MPH+ rockets to the warning track with huge expected wOBA scores. Ian Desmond and Nomar Mazara hauled them in easily, preventing damage. Could smart positioning have played a role?
5. (tie) Aaron Sanchez, Blue Jays
Estimated wOBA: .356
Actual wOBA: .312
Difference: .044
In Sanchez' case, it's a little easier to see what may be the issue. He allowed the same amount of solid contact and barrels as the league average, but he induced less weak contact (5 percent, compared to the average of 7.3 percent) and contact where the hitter got under the ball (13 percent, compared to 21 percent). Those are low-danger batted balls that he missed out on, though he made up for it somewhat with a higher-than-average 46 percent weak grounder rate, compared to 35 percent. It's probably a safe bet that having Kevin Pillar behind him helped, too.

3. John Lackey, Cubs
Estimated wOBA: .378
Actual wOBA: .331
Difference: .046
2. Jacob Arrieta, Cubs
Estimated wOBA: .330
Actual wOBA: .281
Difference: .049
We'll group the two veteran Cubs starters together because the historically good Chicago defense appears to have helped them a similar amount, shaving nearly 50 points off their expected wOBA scores, despite the fact that they were starting from different levels. Arrieta, as expected, was allowing less dangerous contact than Lackey. Jonathan Lester went from .334 expected to .312 actual, a difference of 22 points.
Kyle Hendricks, by the way, had an estimated wOBA of .308, which was the lowest of any qualified National League starter. Remember, that's without considering the aid the Cubs defense actually provided, which suggests that his skill in collecting low-danger contact was a real one. (His actual was .293, which was 15 below his estimated mark.)
1. Seth Lugo, Mets 
Estimated wOBA: .397
Actual wOBA: .295
Difference: .102

Yes, really. You can't imagine the reaction around Statcast™ HQ when we ran this and realized that the King of Curveball Spin was atop a list that had absolutely nothing at all to do with spin rate. That aside, this isn't entirely unexpected, as Lugo's BABIP of .230 was one of the game's lowest and well below the average of .298. He allowed barrels on 10 percent of his balls in play, compared to the average of 6.2 percent.
Even if we step away from batted ball quality to look at the difference between his FIP (4.33, which includes strikeouts and walks) and ERA (2.67), the gap was the second-largest among pitchers with 50 innings. Lugo's curve can be elite, but there are reasons to wonder what he'll offer in 2017.
2016's MOST "UNLUCKY" STARTERS
(Technically, the two starters who topped this list were Timothy Lincecum and Erik Johnson, but Lincecum is currently unsigned for 2017, and Johnson will miss the season due to Tommy John surgery. Let's stick with pitchers you might see this season.)
5. Eddie Butler, Rockies
Estimated wOBA: .409
Actual wOBA: .462
Difference: .053
4. Tyler Anderson, Rockies
Estimated wOBA: .317
Actual wOBA: .377
Difference: .060
Who's surprised that two Colorado pitchers are the first to appear on this list? They are, however, extremely different. While Butler may not have pitched as badly as his superficial numbers seem, even an expected wOBA of .409 is poor, so even if that's what he "should" have had, it's still not great.
Anderson, however, was one of Colorado's bright surprises last year, showing a knack for limiting exit velocity. His expected wOBA of .317 is well above average and matches Rich Hill (keeping in mind this doesn't factor in Hill's strikeout edge). He allowed half as many barrels as the Major League average, and the contact he induced was consistently at or above average. Since his performance against was as expected on the low-velocity hits and well below expected on the harder hits, it seems a safe assumption that the wide expanse of Coors Field hurt him.

3. Jose Berrios, Twins
Estimated wOBA: .420
Actual wOBA: .484
Difference: .064
Berrios was just generally hit hard, as evidenced by a .310/.409/.523 line against. Berrios allowed nearly twice as many barrels as the league average, and while he maybe didn't "deserve" a .484 wOBA, like Butler, the .420 mark isn't impressive either. Even Byron Buxton in the outfield can't help with that.
2. Lance McCullers, Astros
Estimated wOBA: .354
Actual wOBA: .424
Difference: .070
1. Collin McHugh, Astros
Estimated wOBA: .351
Actual wOBA: .426
Difference: .076

Two Astros pitchers were hurt the most by our measure, and it's easy to assume that the dimensions in Houston played a role. It can't just be that, however, because McCullers had the highest BABIP against (.383) of any qualified pitcher since 2012, and home runs aren't counted in BABIP. He did have a well above-average rate of weak grounders (50 percent against the average of 35), but where he got crushed was in "solid contact," where he allowed a .971 wOBA against an expected mark of .670.
Interestingly enough, both McHugh and McCullers were hurt by nearly the exact same amount. Fellow Astros Dallas Keuchel, Doug Fister, and Mike Fiers also underperformed, though not to the same extent. Is it all the ballpark? Something about the Houston defense, either skill or positioning? We can't say for sure, not yet. It does seem Houston isn't a fun place to pitch.