Starting Pitcher Workloads Have Been Significantly Reduced in 2020
During the 2019 postseason, I noted that the period from 1998 to 2015 was fairly stable when it came to starting pitcher usage, even given a wide range in scoring environments and a whole lot of change within the game — expansion, the height of the PED scandal, the advent of PED testing, and the adoption of loose pitch count standards and innings limits on young pitchers. Since that time, we’ve seen starting pitchers throw fewer innings with each passing season, while at the same time generally improving their performance relative to the league.
| Season | IP/GS | Change | K% | BB% | HR/9 | ERA | ERA- | FIP | FIP- |
|---|---|---|---|---|---|---|---|---|---|
| 2015 | 5.81 | -2.61% | 19.5% | 7.1% | 1.06 | 4.10 | 103 | 4.03 | 102 |
| 2016 | 5.65 | -2.8% | 20.2% | 7.7% | 1.24 | 4.34 | 104 | 4.30 | 103 |
| 2017 | 5.51 | -2.4% | 20.6% | 8.1% | 1.34 | 4.49 | 103 | 4.48 | 103 |
| 2018 | 5.36 | -2.8% | 21.6% | 8.0% | 1.21 | 4.19 | 101 | 4.21 | 101 |
| 2019 | 5.18 | -3.4% | 22.3% | 7.7% | 1.44 | 4.54 | 101 | 4.51 | 100 |
| 2020 | 4.73 | -8.7% | 23.2% | 8.6% | 1.28 | 4.13 | 101 | 4.13 | 101 |
Bang! Already what sticks out is that this year, for the first time, starters are averaging fewer than five innings per turn, and their per-start average is down more than one full inning since 2015. What’s more, if this trend continues, it would be the biggest year-to-year drop in innings per start in the span, more than double the drop from 2018-19, and more than triple the other year-to-year drops within that span.
Over the years, a number of factors have driven that decrease, starting with deeper pitch counts, which are a byproduct of higher strikeout rates, as you can see in the table. There’s also the increased understanding of a few sabermetric concepts: starters are generally less effective facing batters for the third time in the game than prior; relievers are generally more effective facing batters for the first time than starters are in any of those appearances; and batters are less effective when they lack the platoon advantage. As starters’ workloads have decreased to account for those factors, their run prevention relative to the league has improved ever so slightly.
The drop-off from 2019 to ’20 is even steeper by a couple other measures:
| Season | IP/GS | Change | TBF/GS | Change | Pitches/GS | Change |
|---|---|---|---|---|---|---|
| 2015 | 5.81 | -2.6% | 24.5 | -2.3% | 93.1 | -2.6% |
| 2016 | 5.65 | -2.8% | 24.1 | -2.0% | 92.6 | -0.6% |
| 2017 | 5.51 | -2.4% | 23.6 | -1.8% | 91.5 | -1.1% |
| 2018 | 5.36 | -2.8% | 22.7 | -3.9% | 88.1 | -3.8% |
| 2019 | 5.18 | -3.4% | 22.1 | -2.6% | 86.4 | -1.9% |
| 2020 | 4.73 | -8.7% | 19.8 | -10.6% | 77.8 | -9.9% |
In terms of batters faced per game, starters are seeing about two fewer relative to last year, and nearly five fewer relative to 2015; again, the scale of the 2019-20 decrease is well beyond that of any of the other year-to-year changes. Likewise for pitch counts.
Of course, it’s early in the season, and we might surmise that out of the gate, starters tend to work less than they do later on in the season, perhaps because even after a full-length spring training, no manager wants to blow out his pitcher’s arm in April, or have him labor too long in the colder weather. Using our custom date range function for each of the past five seasons, I got as close to this year’s total (through August 5) of 332 games started. The pattern is basically the same; this year’s drop-offs are steeper than at any other time in the span, and very similar in size to the full-season ones:
| Season | GS | IP/GS | Change | TBF/GS | Change | Pitches/GS | Change | ERA- | FIP- |
|---|---|---|---|---|---|---|---|---|---|
| 2015 | 338 | 5.72 | -2.1% | 24.2 | -2.4% | 91.3 | -4.1% | 102 | 103 |
| 2016 | 332 | 5.67 | -0.9% | 24.1 | -0.4% | 92.9 | 1.7% | 98 | 96 |
| 2017 | 342 | 5.56 | -1.8% | 23.5 | -2.5% | 90.6 | -2.4% | 92 | 99 |
| 2018 | 324 | 5.38 | -3.3% | 22.9 | -2.4% | 89.5 | -1.2% | 96 | 100 |
| 2019 | 334 | 5.20 | -3.4% | 22.1 | -3.6% | 86.7 | -3.2% | 95 | 98 |
| 2020 | 332 | 4.73 | -9.0% | 19.8 | -10.4% | 77.8 | -10.2% | 101 | 101 |
Honestly, gathering that data was more trouble than it’s worth, as the general outlines of the trend are the same as the full-season data, in that the per-game rates fell from year to year almost without fail. Within a season there’s a bit more movement. Innings per start stabilizes very quickly; the gap between the averages from the first 330-some games and the full season ranged from +0.02 to +0.05 annually from 2016-29, which is to say that they rose slightly, while in 2015, they dropped by 0.09 from the beginning to the end. Batters faced and pitch counts showed more variability in direction and magnitude, but usually moved by less than 1% in either direction. Perhaps of more interest is that the pitchers tended to do a bit better at preventing runs early in the season than they did overall, as indicated by the sub-100 ERA- and FIP-.
Unmistakably, though, the point stands, whether we’re comparing full-season rates with this early stretch, or with similar early stretches from the past half-decade. As a group, starters’ workloads have been cut. They’re down about 9-10% relative to last year, and about a full inning, four batters, and 13 pitches relative to early-season workloads five years ago. That’s a tangible difference.
One factor certainly having an impact upon the starting pitcher workload numbers is the use of openers. The tactic began in 2018 with the Rays, and made its way around the league last year to some degree. Now, there’s no official definition for an opener, and as Jeff Sullivan noted in late 2018 when writing about the team that would soon pluck him away from FanGraphs (sob), when doing bulk research there’s no easy way to select for openers or opener-like appearances, which generally feature a second pitcher — a “primary,” “bulk guy,” or “headliner” — working a longer stint than the opener. Sullivan tracked starts of 40 pitches or fewer, while Baseball-Reference uses two innings or nine batters in its WAR calculations (though it doesn’t publish the number of such outings on its site), and the annual Bill James Handbooks use two innings.
Given that we’ve also seen the stigma against bullpen games, where there may not be any bulk guy, disappear, I think the easiest thing to do is just track the starters by innings and compare them over time. Yes, this will catch the occasional start in which a pitcher who was supposed to go five or more innings gets injured or shellacked and makes an early departure — Max Scherzer left Wednesday’s start after one inning after tweaking his hamstring, and Shohei Ohtani lasted just 1.2 innings in his two starts before a forearm strain effectively ended his season on the mound — but I think this illustrates the trend well enough. Using full-season numbers (and this fragmentary one), here are starts broken down by length:
| Season | <=1 | 1.1-2 | 2.1-3 | 3.1-4 | 4.1-5 | >=5.1 |
|---|---|---|---|---|---|---|
| 2015 | 0.7% | 1.8% | 3.5% | 6.3% | 19.3% | 68.4% |
| 2016 | 0.8% | 2.0% | 3.6% | 7.8% | 21.8% | 64.0% |
| 2017 | 0.8% | 2.0% | 4.4% | 9.4% | 22.6% | 60.8% |
| 2018 | 2.0% | 3.1% | 4.4% | 9.7% | 23.8% | 57.1% |
| 2019 | 3.0% | 4.0% | 5.4% | 10.5% | 24.9% | 52.2% |
| 2020 | 1.8% | 5.4% | 8.7% | 19.3% | 25.6% | 39.2% |
We’ve seen a significant rise in the number of starts of one or two innings in the past three seasons, which coincides with the advent of the opener. Two-inning starts are 2.7 times more frequent now than in 2017, for example. Follow those rows to the right and you can see that starts in the 3.1- to 4-inning range have doubled over the past two seasons, while starts of 5.1 innings or longer are down about 31% in that same timespan.
Indeed, in this young season a number of name-brand starters are carrying much lighter workloads, generally guys working their way back from injuries and/or just not yet built up enough due to the abbreviated summer camps. An incomplete list of recognizable names of those averaging fewer than 5.0 innings per start — excluding the likes of Scherzer, Ohtani, Wade Miley, Corey Kluber, and Mike Soroka, all of whom left starts with obvious injuries — includes Walker Buehler, Michael Fulmer, Zack Greinke, Sean Manaea, Steven Matz, Lance McCullers Jr., Charlie Morton, James Paxton, Rick Porcello, Robbie Ray, Blake Snell, Masahiro Tanaka, and so on.
Back to the data above, here’s a look from a cumulative standpoint:
| Season | <=1 | <=2 | <=3 | <=4 | <=5 | >=5.1 |
|---|---|---|---|---|---|---|
| 2015 | 0.7% | 2.5% | 5.9% | 12.3% | 31.6% | 68.4% |
| 2016 | 0.8% | 2.8% | 6.4% | 14.2% | 36.0% | 64.0% |
| 2017 | 0.8% | 2.7% | 7.1% | 16.6% | 39.2% | 60.8% |
| 2018 | 2.0% | 5.0% | 9.4% | 19.1% | 42.9% | 57.1% |
| 2019 | 3.0% | 7.0% | 12.4% | 22.9% | 47.8% | 52.2% |
| 2020 | 1.8% | 7.2% | 16.0% | 35.2% | 60.8% | 39.2% |
What you’re seeing here is the rate of starts that max out at a given innings total. Starts of two innings or fewer are about as common this year as last year, but those of three or fewer have become more common, to the point that there are more than twice as many this year as there were in 2017 (16.0% versus 7.1%). Last year, less than a quarter of all starts weighed in at four innings or fewer, but this year, that describes more than a third of them, and for the first time, starts of more than five innings are in the minority.
Injuries may well be a factor in these reduced workloads. Even before camps reopened, Luis Severino, Chris Sale and Noah Syndergaard underwent Tommy John surgery. During camp, Marcus Stroman suffered a calf strain, A.J. Puk a shoulder strain, and Jordan Zimmermann a forearm strain, while Eduardo Rodriguez was diagnosed with myocarditis, an apparent aftereffect of his COVID-19 infection. Since the start of the season, Ohtani, Miles Mikolas, and Justin Verlander have all gone down with forearm strains, Kluber has a shoulder strain, and Stephen Strasburg a nerve issue in his hand. The list goes on.
On July 29 at The Athletic, Eno Sarris wrote, “Nearly five times the number of pitchers who went on the injured list in the first week last year have already hit the IL for non-virus reasons this season.” His tally of eight starters who were “sent to the injury list from July 25-29, 2020, compared to days two through five of each season back to 2010” is the highest since at least 2010, and his definition doesn’t even include the aforementioned pitchers who were lost before Opening Day. Thursday’s column by The Athletic‘s Ken Rosenthal included Sarris’ updated count of 20 starters sidelined since the season opened:
An updated comparison from days two through 11 provides additional evidence of the trend, which one executive described as “alarming.” Nearly three times as many pitchers were placed on the IL compared to the same window a year ago, and more than twice as many compared to any of the previous 10 seasons. (The lists exclude IL placements on Opening Day to avoid counting injuries that occurred in training camp, and also excludes placements related to COVID-19.)
As Rosenthal noted, because of larger roster sizes, not every injured pitcher is even on the Injured List. Strasburg was the lone example he offered, but at this writing the Nationals have yet to move Scherzer to the IL as well.
There’s a more comprehensive accounting to be done with regards to injuries –and this deeper dive by The Ringer‘s Ben Lindbergh is worth reading — but it should suffice to say that teams haven’t replaced those pitchers with others from whom they can regularly expect similar length. Minus Sale and Rodriguez, the Red Sox’s rotation is so threadbare that none of the five starters they’ve used besides Nathan Eovaldi and Martin Perez have lasted more than 3.2 innings, thrown more than 74 pitches, or faced more than 20 batters. The Yankees, who not only lost Severino but began the year with Tanaka still recovering from a concussion, are tied with the Giants for the major league lead in most starts of four or fewer innings, a total exacerbated by the early-season struggles of Paxton and J.A. Happ:
The Giants, who are at the other end of the competitive spectrum from the Yankees, have shown signs of disrupting managerial etiquette when it comes to starting pitching, in that new manager Gabe Kapler has taken to a bit of gamesmanship by occasionally delaying the announcement of his starting pitchers as long as possible. On July 25 against the Dodgers, for example, he didn’t announce Logan Webb as his starter until less than four hours before game time, and likewise waited until 90 minutes before first pitch to unveil his lineup. He’s mused about sending two pitchers to the bullpen for pregame warmups, one of them a decoy; doing so with openers would be quite easy. Kapler has used both Tyler Anderson and Kevin Gausman as starters and bulk guys at least once.
It would be an overstatement to say that Kapler has taken this strategy to an extreme, a la the Brewers in the 2018 postseason, when manager Craig Counsell waited until the mornings of the Wild Card game and Division Series opener to announce his starters, prompting Major League Baseball to institute a rule requiring them to do so for postseason games no later than three hours before game time; Counsell also pulled Miley after one batter in Game 5 of the NLCS to confound the Dodgers’ choice of an all-righty lineup. The Athletic‘s Andrew Baggarly, who wrote about Kapler’s disruption, spoke to enough managers to suggest that the day may come when teams keep the information about who’s starting closer to the vest; if openers are involved, the innings and pitch count averages will continue to fall.
Under ordinary circumstances, two weeks would be a blur in a long season, but at this writing, we’re already 19.7% through the schedule. We’ll gain further context on the extent of this trend of starters working shorter as August and September play out, but given concerns about the year-to-year buildups of individual innings totals, the effects of the current situation could continue through next season. By the time the dust finally settles, the expectations for starting pitcher workloads may look quite different.