The Ever-Expanding Postseason Bullpen

The flip side to the story of the Incredible Shrinking Postseason Starter — which, in fact, has only suffered more shrinkage since I published that piece about two weeks ago — is the expanded use of bullpens during this postseason. For the second year in a row, they’re throwing more than half of all innings this October, a situation that probably owes as much to the unique conditions of baseball in the time of the coronavirus pandemic as it does to analytically-inclined management.

Due to a combination of injuries and desperation, this past weekend’s three World Series games saw the six starters combine for just 17.2 innings, with only one lasting longer than four — and that one, Ian Anderson — was removed with a no-hitter in progress! Meanwhile, a total of 28 relievers came through the bullpen gates, an average of nearly five per team per game; they combined for 34.1 innings, nearly double the total of the starters but allowed just 10 runs to the shellshocked starters’ 24. Yes, the Braves’ use of the inexperienced Dylan Lee and Tucker Davidson to start Games 4 and 5 had something to do with it, but even Framber Valdez, the author of this postseason’s only eight-inning start, was chased after allowing five runs for the second straight turn. Valdez, in fact, has allowed 10 runs in this World Series, while the rest of his staffmates have yielded just eight.

Differentials like that, in which relievers have been much more effective than starters, help to explain what we’re seeing. While starting pitchers are often the stars who offer more name recognition to casual fans, given the urgency of October baseball there’s little sense in riding a starter who’s getting beaten up, regardless of the underlying reasons or the theoretical entertainment value. With the Astros carrying 13 pitchers on their 26-man World Series roster and the Braves 12, there are more than enough relievers to go around. Just one reliever pitched in Games 3, 4, and 5 — which is to say, made back-to-back appearances without a day off: the Astros’ Ryne Stanek, whose first appearance was just one batter and four pitches long. Only one other reliever in the entire postseason has pitched on three straight days, the Red Sox’s Martín Pérez.

Beyond the ample supply of bullpen arms, we’re seeing the amplification of trends that we’ve observed during recent regular seasons. The relievers as a group are more effective at preventing runs than the starters, albeit not by a great deal:

Regular Season Starter vs. Reliever Comparison
2015 4.10 3.71 -0.39 4.03 3.83 -0.20
2016 4.34 3.93 -0.41 4.30 3.99 -0.31
2017 4.49 4.15 -0.34 4.48 4.16 -0.32
2018 4.19 4.08 -0.11 4.21 4.06 -0.15
2019 4.54 4.46 -0.08 4.51 4.51 0.00
2020 4.46 4.44 -0.02 4.46 4.45 -0.01
2021 4.34 4.17 -0.17 4.30 4.22 -0.08
RP Adv. = reliever advantage (differential relative to starter ERA or FIP)

The advantage typically grows in the postseason, even when the innings are distributed more heavily among the better pitchers while the fifth starters and lower-run relievers are dispensed with:

Postseason Starter vs. Reliever Comparison
2015 4.33 3.55 -0.78 4.17 3.40 -0.78
2016 3.88 2.88 -1.00 3.86 3.16 -0.70
2017 4.08 3.97 -0.11 4.32 4.20 -0.12
2018 3.90 3.60 -0.31 3.96 3.75 -0.22
2019 3.36 4.24 0.87 3.91 4.23 0.32
2020 4.25 4.06 -0.19 4.55 4.32 -0.23
2021 4.65 3.93 -0.72 4.04 3.82 -0.22
RP Adv. = reliever advantage (differential relative to starter ERA or FIP)

For the seven-season period, relievers have a 0.22 runs per nine advantage during the regular season in terms of ERA (4.12 to 4.34) and a 0.28 runs per nine advantage during the postseason (3.78 to 4.06), but this year the gap is about two-and-a-half times that. Generally, the gap is because starters give up more homers and strike out fewer hitters than relievers, though this year, it’s the latter giving up more homers (1.31 to 1.23 per nine) but still having an edge in both BABIP (.298 to .306) and quality of contact (.373 xwOBAcon to .397).

Thus, we’re seeing more relievers per game but a shift to fewer outs per reliever….

Regular vs. Postseason Reliever Usage
YR Reg RP/G Reg IP/App Reg IP/G Post RP/G Post IP/App Post IP/G
2015 3.11 1.01 3.13 3.42 1.05 3.59
2016 3.15 1.04 3.27 3.64 1.07 3.89
2017 3.22 1.05 3.39 3.61 1.14 4.12
2018 3.36 1.07 3.58 4.42 1.05 4.63
2019 3.41 1.10 3.76 3.95 0.95 3.76
2020 3.43 1.12 3.83 4.09 1.08 4.42
2021 3.43 1.09 3.75 4.46 1.09 4.87
RP/G and IP/G are relievers used and innings pitched per team game. IP/App is innings per reliever appearance.

…and a greater share of innings thrown by relievers, both as a regular season trend in general and an even sharper increase in the postseason:

Regular vs. Postseason Innings Shares 2015-21
Year SP IP/G Reg RP IP/G Reg %RP Reg SP IP/G Post RP IP/G Post %RP Post
2015 5.81 3.13 35.0% 5.51 3.59 39.5%
2016 5.65 3.27 36.7% 5.11 3.89 43.2%
2017 5.51 3.39 38.1% 4.73 4.12 46.5%
2018 5.36 3.58 40.1% 4.68 4.63 49.7%
2019 5.18 3.76 42.1% 5.14 3.76 42.3%
2020 4.78 3.83 44.5% 4.40 4.42 50.1%
2021 5.02 3.75 42.7% 3.95 4.87 55.2%

That trend — which I’m expressing as a percentage of innings by bullpens instead of by starters, which I did in my previous exploration — owes something to expanded bullpens. The two highest shares of regular season innings have come in the past two seasons, which were played with larger rosters than the 25-man limit that was in place through the end of the 2019 season (excluding the extra player for doubleheaders, introduced in 2012). Because of last year’s shortened summer camps, teams were allowed to carry 30 players for the first two weeks of the season and then 28 for the remainder, including the postseason. This year, roster limits were at 26 for both the regular season and postseason, and while the number of active players in September has been reined in (from 40 to 28), postseason rosters are still larger than they were prior to 2019, and it has resulted in more pitching changes.

We’ve seen more changes in both the regular season and postseason despite the introduction of the three-batter minimum, which was supposed to cut down on the number of mid-inning pitching changes and speed up games. The rule is applicable for starters and relievers, though a pitcher who finishes an inning can be removed regardless of how many batters faced. It has cut down on the number of one- and two-batter appearances in both the regular and postseasons (though we still see more of the latter in the playoffs than during the rest of the year), but the average time of nine-inning games has gotten longer nonetheless:

Regular vs. Postseason Short Relief Stint and Time of Game
Year Reg 1-2 BFP Post 1-2 BFP Reg Time Post Time
2015 17.1% 11.0% 2:56 3:20
2016 14.8% 14.1% 3:00 3:29
2017 14.2% 19.0% 3:05 3:40
2018 14.1% 12.3% 3:00 3:40
2019 13.1% 9.6% 3:05 3:40
2020 8.5% 12.4% 3:07 3:38
2021 8.2% 11.8% 3:10 3:45
SOURCE: Baseball-Reference
Percentages of reliever outings facing either one or two batters (3-batter minimum rule in effect for 2020-21). Times are per nine innings.

More pitching changes! Longer games! Less familiarity with the pitchers because the starters are making early exits! The average start is now down to 3.95 innings, 0.45 before last year’s all-time low. Who doesn’t love all that? A whole lot of us, judging by the tenor of conversations on social media, and it’s understandable. Right now, postseason baseball feels very different from the regular season, and more disorienting for those of us who can remember starting pitchers going the distance once in awhile during the postseason (the last complete game was Justin Verlander’s 13-strikeout effort in Game 2 of the 2017 ALCS, and 2016 is the only postseason with more than two in the past decade).

For as frustrating as this may be to watch at times, I’m still falling back on the explanation of the starters’ workload increases in the wake of the coronavirus pandemic as the reason we’ve seen managers hand the keys to the bullpen even more quickly than usual. The four teams that played in the LCS featured 13 pitchers with year-over-year increases of at least 90 innings, and most of those pitchers have been pretty spotty in October, unable to string together two strong outings in a row and now being kept on short leashes. Picking up a table I put together for my coverage of the Anderson start (it ran on Saturday, so many readers may have missed it):

2020-21 Workload Increases Among LCS Pitchers
Pitcher Team 2020 Reg 2020 Post 2021 Reg 2021 Post Increase
Walker Buehler LAD 36.7 25.0 207.7 18.3 164.3
Nathan Eovaldi BOS 48.3 0.0 182.3 20.7 154.7
Luis Garcia HOU 12.3 2.0 155.3 13.0 154.0
Nick Pivetta BOS 15.7 0.0 155.0 13.7 153.0
Charlie Morton ATL 38.0 20.0 185.7 16.7 144.3
Max Scherzer LAD 67.3 0.0 179.3 16.7 128.7
Julio Urías LAD 55.0 23.0 185.7 15.0 122.7
Ian Anderson ATL 32.3 18.7 143.0* 17.0 109.0
Max Fried ATL 56.0 23.7 165.7 21.7 107.7
Drew Smyly ATL 26.3 0.0 126.7 4.3 104.7
Lance McCullers Jr. HOU 55.0 14.7 162.3 10.7 103.3
Jake Odorizzi HOU 13.7 0.0 104.7 6.3 97.3
Zack Greinke HOU 67.0 14.7 171.0 2.3 91.7
*includes 14.2 minor league innings via rehab assignment.

Will things return to normal? What is normal anyway? I don’t have the answer to that, and so long as teams are aware of the third-time-through-the-order penalty, I think we’ll continue to see a lot of postseason starts fall short of five innings. But I also think that what we’re seeing right now is not the new normal, it’s a reaction to unique, subpoptimal conditions for starters. Check back next year.

Brooklyn-based Jay Jaffe is a senior writer for FanGraphs, the author of The Cooperstown Casebook (Thomas Dunne Books, 2017) and the creator of the JAWS (Jaffe WAR Score) metric for Hall of Fame analysis. He founded the Futility Infielder website (2001), was a columnist for Baseball Prospectus (2005-2012) and a contributing writer for Sports Illustrated (2012-2018). He has been a recurring guest on MLB Network and a member of the BBWAA since 2011. Follow him on Twitter @jay_jaffe.

newest oldest most voted

Enjoyed the article. I imagine it’d be too cumbersome to calculate, but I wonder whether considering the second or “bulk” pitcher in a bullpen game (or maybe whichever of the first two pitches more innings) to be the “starter” would contribute to any insight into what’s happening with the innings of the “real” relievers. Seems like bulk pitchers would misleadingly inflate the IP/appearance of relievers.