Fister and Regression

Doug Fister carved out a career as a high quality starting pitcher in Seattle and Detroit from 2009 - 2013, with an average  ERA+ of 115.  After he was traded to the Nationals, he had a terrific 2014 season, with an ERA of 2.41.  At the age of 31, he had a disappointing season before hitting free agency. In 2015, He produced a 4.19 ERA (95 ERA+) and 4.55 FIP for the Nationals, far worse than his normal results.  At the age of 32, he entered free agency coming off his worst season, with teams wondering if his decline would get worse as he aged.  This created a potential win-win situation for he Astros and Fister.  He signed a one year $7 million contract; and, from his perspective, this would permit him to rebuild his value.

So far, Fister has produced a 3.91 ERA (101 ERA+) and 1.4 WAR for the Astros.  Taking into account that he already has provided the Astros just under one and a half wins above replacement, the Fister signing can be considered a good investment by the Astros.  But the Astros are in the stretch run of contending for a playoff slot, which raises the question, can Fister continue to pitch well for the Astros over the next month and into the playoffs?  Moreover, with the Astros' ace, Dallas Keuchel, out with shoulder inflammation, Fister's future performance becomes even more critical.

Regression

You might hear two contrasting views from fans.  Some will point to the battering he took in his last start against the Rangers and say that he can't be trusted.  Others will point to his 3.91 ERA and say that Fister has been better than Keuchel anyway--so why worry? Neither reaction is satisfactory.  As for the Rangers' loss, one game is a tiny sample size, and really can't be used to predict much of anything.   And, although ERA describes past results during the year, the ERA over the course of a single season is a weak predictor of future performance.  Fielding Independent Pitching (FIP) is based on peripheral pitching stats most directly under the pitcher's control, and is generally more predictive of future ERA than the pitcher's current ERA.

And this is where predicting Fister's future performance becomes more complicated.  For most of this season, Fister's FIP has been running more than 1 run worse than his ERA.  This brings up the "R" word--regression.   A large deviation between FIP and ERA implies that ERA is likely to regress in the direction of FIP.

An additional complicating factor is that some pitchers seem to be capable of out-pitching their FIP on a regular basis--although a differential of more than 1 run probably isn't sustainable over a large enough sample.  And. over the course of his eight year career, Fister has shown the ability to maintain an ERA somewhat below his FIP (career ERA of 3.49 and career FIP of 3.74).

Team defense can be another complicating factor.  The "fielding independent" part of FIP means exactly that; FIP purposely excludes defense.  And, good or bad team defense can account for variation between FIP and ERA.  Advanced defensive metrics are not without controversy.  But if you believe that the Fielding Bible DRS (defensive runs saved) metric is accurate, then Fister's FIP should be higher than his ERA.  DRS shows the Astros team defense as the best in the AL, by a wide margin.  Including runs saved by the Astros' constant shifting, the Astros' defense has saved 80 runs above average (i.e.,approximately nine wins).  This is 24 runs saved more than than the next best AL team (Blue Jays).  Of course, knowing that Fister's ERA should be lower than his FIP does not tell us how much lower.   Shortly after Fister was signed, Fangraphs suggested that Fister would benefit by moving from the team with the fewest shifts (Nationals) to the team with the most sophisticated defensive shifting (Astros) ("The Case for Doug Fister" by Scott Spratt, Feb. 4, 2016).

Has Fister Begun to Regress?

Statistical principles refer to "regression to the mean."  What average constitutes the "mean?"  This is usually represented as regression to the population mean, or in this case, league average.  However, sometimes baseball analysts may express regression relative to the player's expected talent level, as indicated by projections or career average performance.  The table below shows comparative targets for both forms of regression: the AL average starting pitcher in 2016, and the fangraphs composite projection for Doug Fister (referred to as depth chart projection).

The difference between Fister's first half and second half performance, below, supports the idea that Fister's performance has begun the process of regression.

Interestingly both ERA and FIP appear to be regressing, which entails movement in opposite directions. Sometimes we make assumptions about the effect of regression on a single statistic without recognizing that multiple statistics are regressing at the same time.  In the second half, Fister's ERA regresses in the expected direction toward his higher FIP, but at the same time his FIP regressed downward in the direction of the mean.  To the extent this is completely explained by regression, the second half movement has overshot the mean (either AL average or Fister projection) with the ERA and BABIP higher than the target and FIP lower than the target.  This isn't unusual, since regression usually doesn't follow a straight line, but it doesn't provide much help figuring out where Fister is headed this season.

Without a doubt, Fister's results from the most recent Rangers' series had a significant effect on his second half ERA.  To some extent, this supposed second half "regression" could reflect a string of games against teams with heavy lefthanded batting lineups.  Three of Fister's blow outs in the second half were against teams with quality lefthanded bats (Yankees, Rangers, and Rays).  Doug Fister has significant splits against lefthanded hitters.  Specifically, he has been hit hard and often by lefties: LH bats have a .894 OPS and RH bats have a .583 OPS.  If Fister could pitch only against righthanders, he might be a Cy Young candidate, since they only have a .214, .276, .307 slash line against him. Over the rest of season, watch the number and quality of LHBs who will face Fister.

Possible Explanations for ERA Less Than FIP

As mentioned above, Fister has shown a history of maintaining an ERA below his FIP.  (In some ways, this reminds me of Scott Feldman.)  The differential is modest (about one quarter run).  This suggests to me that, by season end, Fister's ERA may not regress upward all the way to league average.

What could explain this ability?  I have already mentioned the Astros' defense which may turn some of the sharp liners and gappers into outs.  The Fangraphs article cited previously pointed out that Fister had the second highest BABIP on groundballs (.304) among all 2015 starters and noted that the Astros had the third lowest BABIP on groundballs (.215) in 2015.  Fister has a very good .203 BABIP on groundballs with the Astros this year.  (.248 is league average.)  Some of this BABIP suppression on groundballs may be defense, some may be luck, and...don't forget soft contact.

Some pitchers may have an inherent ability to induce soft contact, whether due to deception, sequencing, or movement.  And some analysts contend that such "contact management" skill enables those pitchers to suppress their ERA below FIP.  Among starting pitchers, Fister is 24th best at soft contact and 32d best at avoiding hard contact.  This may not be elite contact management skill, but it is well above average.

About a month ago, on the writer's listserve, Ashitaka mentioned that Fister had one of the lowest swing rates on pitches in the strike zone.  I have noticed since then that Fister continues to be at the top of that list.  It's unclear if this is a skill that enable him to outperform FIP.  Fister keeps interesting company on this ranking: 1 through 5 in order of lowest swing % in-zone are Kyle Hendricks, Doug Fister, Jose Fernandez, Adam Wainwright, and Tanner Roark. One of these pitchers (Fernandez) is not like the others, because he is a high velocity, high strike out pitcher.  The others are similar to Fister in that they are not pitchers with high velocity and pair a low strike out rate with an exceptionally good walk rate.  Hendricks and Roark are 1 and 2 on the soft contact ranking.   Hendricks has been mentioned as a Cy Young candidate in the NL, and probably has pitched better than his more acclaimed teammate, Jake Arrieta.  Hendricks is successful--despite velocity in the 88 mph range-- because (among other things) he has a good change up and his command is extremely precise.

I am guessing that a low swing rate on pitches in the strike zone is associated with the pitcher's ability to locate his pitches on the edges and/or surprise hitters with good sequencing of breaking pitches.  Is this a repeatable skill?  I don't know, but if my explanation is correct, it could be.  And, by the way, Fister had similarly low swing rates in-zone while pitching in Detroit.

Negative Trends

Besides Fister's rising ERA and BABIP in the second half, several other negative impacts stand out.  Fister's groundball rate is much lower in the second half (49% first half and 38% second half).  I don't view the second half groundball rate as sustainable because it varies significantly from his career profile.  Most likely this is a small sample aberration, and Fister's groundball rate will regress in the direction of his career average (48%).  In addition, Fister's second half HR/Fly rate (6.5%) is significantly lower than the first half.  It may not seem like this is a negative.  But it may be difficult to keep the HR/fly rate that low in the future.  The low second half HR/fly rate contributed to the low FIP in the second half.

Conclusion

I don't feel like I am much closer to predicting Fister's performance over the upcoming September and October.  Fister's performance is regressing, but this does not mean that he is likely to collapse.  Fister's projections indicate that his ERA and FIP should be similar to an average starting pitcher. In my view, the most likely expectation over the remainder of the season should be "average starting pitcher."  That could mean a No. 3 or 4 rotation performance. He could produce results like a top of the rotation pitcher, though it's not likely.  But average starting pitcher performance is not really a bad thing.  If the offense can produce well, and the defense provides Fister with good support, Fister has the capability of keeping the team in games.