Michael Pineda was bad last year. Although he finally stayed healthy for a full season, the results were disappointing. Pineda pitched only 175 innings in 32 starts, posting a 4.82 ERA. Famously (now at least), he led the American League in strikeout rate, and posted a solid 3.79 FIP. and stellar 3.30 xFIP. The gulf between his ERA and FIP/xFIP led to quite a bit of disagreement about how good Pineda was in WAR measurements.
How good will Pineda be next year? I think he's a good bet to be very good, for three reasons.
His ERA Overstates How Bad He Was Last Year
It sure felt like Pineda was giving up a lot of runs last year. 4.82 may not have seemed that high at other points in history, but it felt high last year. This in part due to the more offensive run environment that AL starters faced in 2016. As recently as the 2014 season, the average AL starter posted a 3.92 ERA. In 2016, the average AL starting pitcher posted a 4.42 ERA. Pineda was 9% worse than league average. That's bad, but not horrible.
The ordering of events made Pineda seem worse as well. He started off the season very poorly with an ERA peaking at 6.92 in late May. It came down as he improved, including doing some of his best work in September.
He Was Probably Very Unlucky Last Season
The ERA thing was pretty clear and uncontroversial. This is a little more debateable.
Pineda's very good 3.79 FIP masks what could have been an even better season. The theory behind using a metric like FIP involves crediting both the defense and pitcher for defensive contributions or failures. ERA gives 100% of the credit to the pitcher on the mound for defense. We all know intuitively that this is not true. The defenders on the field deserve a lot of credit. Luck and other exogenous factors (like umpires) do as well. The criticism of FIP is that it is too simplistic. A lot more happens on the field that is to some degree under the control of the pitcher than just strikeouts, walks, and home runs.
Luckily, we have better statistics than FIP. Baseball Prospectus puts out a statistic called Deserved Runs Average. Think of it as FIP on steroids. I highly recommend this primer on DRA. I won't explain everything here, but to give you an idea of how thorough it is, DRA controls for (among other things): park factors, weather, strength of opposition, framing, controlling the running game, sequencing, relief pitcher inherited runners, and passed balls.
That's a lot! These components and more are fed into a model of three seasons of baseball, then used to estimate the performance of a pitcher. Since we have a really clear outcome measure (runs allowed), we can train a model to do a very good job predicting the future. DRA is really good.
Who were the top players in the MLB by DRA last season?
What!? Michael Pineda was the best pitcher in the American League on a per-inning basis last season? Yeah, I don't believe it either. There is a lot of face validity to the rankings. Jose Fernandez (RIP), Chris Sale, Justin Verlander and Noah Syndergaard were definitely some of the starting pitchers in the majors last season. However, Pineda is not the only name that stands out on the list. Robbie Ray is the other big one.
Here's what I believe is going on: Pineda was both unlucky and the system is missing something about his performance. There's a lot of ways that pitchers can be unlucky that we don't normally think of. Relief pitchers can allow a lot of extra inherited runners to score. They could end up drawing the tough matchups or away ballparks. They could pitch on nights with bad weather conditions, etc.
There's also a lot of ways pitchers can pitch worse than a FIP metric might predict. They could allow hard contact in bunches, rather than independently. They could allow more hard contact in general than the average pitcher.
My bet is that there is some signal and some noise in Pineda's incredibly DRA. Pineda is a good bet to be much better than 2016, but not a good bet to be as good as his DRA might predict.
Pineda Still Has a Lot of Talent
I think one factor that is easy to forget about is how good Michael Pineda's stuff is. His fastball averaged 94 mph last season, 13th best in the majors. His slider graded out as 17th best in the majors as well. His command and control aren't bad either. More broadly, Pineda was until fairly recently considered one of the more talented young pitchers in the league, derailed more by injury than anything else. Jesus Montero was a big deal when we traded him for Pineda.
Talent obviously does not automatically lead to performance. However, statistical theory suggests we shouldn't ignore it. Pineda is still relatively new to the major leagues, even though it feels like he has been around forever. There is a sphere is statistics called Bayesian. Without boring you too much, Bayesian statistics make predicts the future based upon both new information (Pineda's pitching outcomes) and our prior expectations about how good Pineda is at pitching. When we receive new information, we update those expectations, but don't completely throw them out.
There's statistical reasons to believe Pineda is actually pretty good (possibly great). There are plausible reasons to believe he has the physical ability to be pretty good (possibly great). His performance (measured by ERA at least) has been poor over the last two seasons. Given the first two, there is some significant probability that the poor performance is caused by random chance, rather than poor talent for baseball.