The death of the starter has been greatly exaggerated. On the heels of a regular season in which starting pitchers threw a smaller share of innings than ever before, and one year after a postseason in which they threw barely over half the total innings, it seemed quite possible that the trend might continue this October, particularly with each of the four 100-win teams spending September scrambling for a fourth option and some of them publicly floating novel ideas about how things might unfold. Admittedly, it’s early in the 2019 postseason, but already we’ve seen some monster pitching performances in the playoffs — Justin Verlander‘s seven innings of one-hit ball, Gerrit Cole‘s 15 strikeouts, Adam Wainwright’s 7.2 scoreless innings — and in general more reliance upon teams’ front-of-the-rotation starters than in the recent past.
Consider: in all of the 2018 postseason, just four times did a starting pitcher throw 100 pitches in a game: Walker Buehler twice (in Games 3 of the NLCS and World Series), Hyun-Jin Ryu once (Game 1 of the NLDS), and Verlander once (Game 1 of the ALDS). Already this year, seven pitchers have done it: Buehler (again), Verlander (again), and Patrick Corbin in their respective Division Series openers, Cole and Jack Flaherty in their respective Games 2, Adam Wainwright in Game 3, and Max Scherzer in Game 4. The count of seven-inning starts isn’t quite there yet, but last year, there were nine in 66 total starts (13.6%), but just four in the Division Series. This year, there have been seven in 28 Division Series starts (25%):
|Justin Verlander||ALDS Gm 1||HOU||TBR||7.0||1||0||0||3||8||0||100|
|Gerrit Cole||ALDS Gm 2||HOU||TBR||7.2||4||0||0||1||15||0||118|
|Jack Flaherty||NLDS Gm 2||STL||ATL||7.0||8||3||3||1||8||1||117|
|Mike Foltynewicz||NLDS Gm 2||ATL||STL||7.0||3||0||0||0||7||0||81|
|Adam Wainwright||NLDS Gm 3||STL||ATL||7.2||4||0||0||2||8||0||120|
|Mike Soroka||NLDS Gm 3||ATL||STL||7.0||2||1||1||0||7||0||90|
|Max Scherzer||NLDS Gm 4||WSN||LAD||7.0||4||1||1||3||7||1||109|
As we’re in small-sample territory, this may well be randomness at work. More likely, it’s a byproduct of these teams having time to set up their rotations down the stretch, and working in a bit of extra rest along the way. Either way, it bears watching over the next few weeks given the backdrop of the regular season. In 2018, the share of innings thrown by starters dropped below 60% for the first time (59.9%), and this year, it fell by another two points — 903 fewer innings thrown by starters. The total number of innings thrown by starters is down by 3,834.1 relative to 2014, just five years ago. Here’s a graph of it by volume:
As you can see, things were fairly stable from 1998-2015, that despite a very wide range in scoring environments, from as high as 5.14 runs per team per game in 2000 to as low as 4.07 in ’13. During that span, which included expansion, the height of the PED scandal, the advent of testing, and the adoption of loose pitch count standards and innings limits on young pitchers — basically, a whole lot of change. And yet, the major league-wide average innings total for starters (2,8685.1) remained in a narrow range, with a standard deviation of just 381 innings per year, about 1.3% of the total. The total workload fell below 28,000 innings for the first time in 2016, then below 27,000 in ’17, and below 26,000 this year.
The primary drivers of the drop, I believe, are deeper pitch counts — a byproduct of higher strikeout rates — and an increasing understanding that starters are generally less effective in facing batters for the third time in the game than prior, that relievers are generally more effective in facing batters for the first time than starters are in any of those appearances, and that batters are less effective when they lack the platoon advantage. Hence the increasing reliever workloads. Here’s a look at the average number of batters faced per game by starters, broken down by times through the order. Obviously, the maximum is 9.0, and thanks to the rise of the opener, we’ve seen the average for the first time fall from 8.97 as recently as 2017 to 8.85 this year, with the average for the second time falling from 8.71 to 8.24 in that span, and the third time falling from 6.68 in 2014, to 5.99 in 2016 to 5.13 last year and 4.84 this year:
Overall, the volume of batters faced by pitchers a third time through the order has dropped 23.3% since 2015. Over the past two years, we’ve seen an interesting change in the performance patterns as well:
Last year, for the first time during the time frame covered by our splits (2002 onward), relievers were less effective the first time they faced a batter than starters were. This tracks with Craig Edwards’ observation from August 21 that relievers were no longer out-pitching starters in terms of FIP. While the gap was as big as 0.32 runs per nine just three years ago (starters 4.48, relievers 4.16), it was cut to 0.15 last year (starters 4.21, relievers 4.06) and disappeared this year (4.51 for both). In that time, scoring levels have risen again, the strategy of using openers has gained some popularity, there’s been an increasing polarization between good teams and bad, and possibly some talent dilution as well.
That part of things is an issue for another day. The point is, starter regular season workloads: down. In recent years, we’ve seen it drop to an even greater degree in the postseason…
…except for this year, or at least the part of this year that’s through Monday (both Wild Card games, plus four games apiece in the NLDS, three apiece in the ALDS) — 32 total starts. We’re seeing around a seven-point bump over last year’s level, from 50.3% to 57.6%, or 59.0% if you use Division Series Game 3 as a cutoff for the purposes of year-to-year comparisons; again it’s worth noting that we’re talking mainly about the fronts of rotations, aces and generally-solid second and third starters. And they’ve pitched well, posting a 2.73 ERA, down from last year’s overall mark of 3.90 and thru-DS Game 3 mark of 3.47. If this ERA holds (ha!) it would be the lowest mark since 2001 (2.69). Meanwhile, bullpens have been lit for a 5.14 ERA through Game 3, which would be the worst mark of the Wild Card era by nearly half a run (2004, 4.65). With every bullpen performing more or less like that of the Nationals during the regular season — they were 24% above the major league average scoring level in terms of runs per nine, while all relievers right now are 26.6% above — it’s no wonder managers are sticking with starters.
Workload-wise, it’s worth comparing where we are to this point, relative to the recent past. Dialing back to the start of the Wild Card game era in 2012, since that has an effect on the choice of starting pitchers:
As you can see, the early-postseason share of starter innings has varied relative to the overall share — some years lower, some higher — but this year is as high as it’s been since 2015, the year that marks the endpoint of the long stretch of regular season stability. Perhaps that share owes something to the talent level of the pitchers in question, and while there are multiple ways to determine that, a spot check suggests there’s something to that theory. Through all of the Division Series Games 3 this year (28 starts total), 15 were taken by pitchers with a FIP- in the majors’ top 30, while 14 were taken by pitchers with an ERA- in the top 30 (120 IP cutoff for both). Using the same specs, last year, when the share was 8.4 percentage points lower, the counts were 12 by FIP- and 13 by ERA-, while in 2015, before things started to change, it was 15 by FIP- and 14 by ERA- — the same as this year.
All of this is rather striking given the chatter we heard in September, as even the best teams scrambled for a fourth postseason starter and floated ideas suggesting departures from orthodoxy. Even before they lost Domingo Germán to administrative leave in connection with a domestic violence allegation, Yankees manager Aaron Boone told Sports Illustrated’s Tom Verducci, “We’re going to be a little untraditional. The only one we might use as a traditional starter is Paxton.” Boone envisioned shorter stints for Luis Severino, if he proved to be healthy upon returning from a lat strain, with Germán and/or J.A. Happ in piggyback roles, and Chad Green possibly serving as an opener. Masahiro Tanaka even took his final turn of the season by coming out of the bullpen following Green, “in case we ever were to do something where it’s a little bit of a unique role or doing an opener role,” as Boone explained.
Elsewhere, Rich Hill‘s mid-September MCL sprain prevented him from getting stretched out to handle a full starter workload and threw a handful of other Dodgers — none of them stretched out either — into the mix for their fourth spot. Michael Pineda’s PED suspension and the rough second halves of Kyle Gibson and Martín Pérez led the Twins to start rookie ex-Uber driver Randy Dobnak, who had opened and started a total of five times all season, in Game 2 against the Yankees, with the number four spot up for grabs. Neither Hill nor Dobnak survived the third inning of their respective games, bringing the starter innings share stats down somewhat. As the Twins were swept, we’ll never know just what Rocco Baldelli had up his sleeve for Game 4.
Through Monday, none of what unfolded, starting-wise, did much to push the envelope unless you count the Braves’ Dallas Keuchel pitching on three days of rest while the team held Mike Foltynewicz back for Game 5. On Tuesday, the Astros will send Verlander back to the mound on three days’ rest; Wade Miley’s struggles over his final six starts (12.60 ERA, 6.08 FIP) cost him the fourth starter job, and the team chose to bypass 24-year-old rookie Jose Urquidy, who has just 41 major league innings under his belt. The Rays will be this year’s first team to detour from convention — because of course they will — countering Verlander with Diego Castillo, who’s never gone more than two innings in 17 turns as an opener, starting as part of a bullpen game.
With the Yankees moving on to a best-of-seven series, we’ll see where they go with their fourth starter; as for the other series, things are still up for grabs. Having tracked the numbers underlying this trend for a couple of days before publishing this, I imagine the overall performance levels will up more closely aligned with the recent past, but for now, the contrast is striking enough for the trend to bear watching.
var SERVER_DATA = Object.assign(SERVER_DATA || );