Last year, the 102-win Cleveland Indians were one of the great teams of the modern era: Their starting pitchers had the American League's lowest ERA, their offense scored more runs per game than all but two AL teams, and their defense was, by Defensive Runs Saved, the AL's third best.
And their bullpen! Collectively, Cleveland's relievers had the game's lowest ERA. Their sOPS+ -- OPS allowed, relative to the rest of the league's relievers -- was baseball's 19th best since 1988, when the current era of bullpen usage more or less began. The Indians' bullpen was so good and so deep that the club had to leave three relievers with ERAs in the 2s off the postseason roster.
This year, the club is fine. The Indians are playing at an 86-win pace, which is just good enough to feel safe in one of the worst divisions in recent memory. But it's nothing like last year's juggernaut, despite very good starting pitching (second-best ERA in the American League), a very good offense (more runs per game than all but three AL teams) and a very good defense (the AL's third best by DRS).
But their bullpen! Collectively, Cleveland's relievers have the game's highest ERA. By sOPS+, they have the seventh-worst bullpen since 1988. It's enough to make you wonder why Cleveland's front office traded away all the relievers over the winter and replaced them with much worse relievers, but, of course, the Indians didn't do any such thing:
Free agency cost the Indians two key members of last year's bullpen: Joe Smith, who was acquired on July 31 and made the postseason roster, and Bryan Shaw, who led the club in appearances. Clearly, Cleveland is reeling from the losses of those two reliable veter -- whoops-a-daisy, Smith's ERA for Houston is 5.49, and Shaw's in Colorado is 7.09. A few other guys pitched scattered innings last year, and a few others have pitched this year, but the bulk of relief work was or has been handled by the names we just listed.
So let's just recap: Cleveland had one of the best bullpens ever last season. The Indians brought most of it back. They have one of the worst bullpens ever this season. We have identified what might be the cruelest part of the modern game.
There's a case that sports generally are tilting more toward high-variance strategies. The fly ball, for instance, is a high-variance offensive weapon: Fly balls that stay in the yard are more likely than any other ball to be turned into an out, but fly balls that clear the wall are the most valuable act in the sport. The margin between the two can be impossibly thin over the course of a single game. Two teams might play identically, but one could outscore the other by six on a couple of big flies alone. In basketball, 3-pointers are a high-variance strategy, as is a pass-heavy offense in football. The analytics support these strategies, but when they fail, as they inevitably will at times, a good team looks temporarily awful.
Major league baseball teams have, by turning more and more innings over to bullpens, collectively adopted a high-variance strategy that can make a good team look bad not for a game or a week but for a season. It's almost certainly wise to lean heavily on relievers, the better of whom are generally more dominant in their one-inning bursts than (even superior) starters are over seven or eight innings. Relievers can be matched up against batters so the team on defense has the platoon advantage more often. Relievers can be saved until the most perilous moments so the best pitchers' finite pitches aren't wasted in six-run games. It's good strategy, which is why it's what the game has been moving toward since, oh, 1950? Maybe earlier.
This year, relievers will set a record for innings pitched -- a record broken at least a dozen times in the past 30 years and a record that'll probably be broken a dozen times more:
It isn't impossible to win with an ineffective bullpen, but it's probably less possible than it used to be, when starters bore more of the burden and when clubs needed only a couple of relief aces to handle the late-and-close, high-leverage assignments.
But the tradeoff to this sound strategy is that relievers, well, almost two-thirds of the closers lose their jobs over the course of a single season, so you can imagine what it's like further down the assembly line. Relievers are like slices of buttered bread constantly falling off tables, and it's anybody's guess whether they'll land with the butter side up. When they do, you're in the playoffs. When they don't, for whatever reason -- small-sample flukes, a few weeks of wildness, an unexpected injury, an expected injury, age-related velocity declines, loss of feel for one pitch or the fact that if they're relievers they maybe weren't that good to begin with -- you've got a carpet covered in butter at the worst possible time.
Cleveland's collapse-by-committee is an extreme example, but this sort of thing is happening constantly. From one year to the next, a team's bullpen performance fluctuates so much that some years there appears to be no carryover at all. When a team does make big moves, the results often seem almost random: The Rockies spent nine figures building a "super bullpen" last winter, but their bullpen's Win Probability Added has actually plummeted, from third in the league last year to 16th this year. Or the reverse: The A's sold off their two best relievers at last year's trade deadline, yet their atrocious 2017 bullpen has morphed into baseball's third best this season.
Of course, teams will change players, and players will change career trajectories, no matter the position or role. But we can measure the variance.
From one year to the next, team defense -- as measured by Ultimate Zone Rating -- has shown a correlation of about .44 since 2013. (In statistics, a correlation of 1 means a perfect correlation, and a correlation of 0 means no more relationship than randomness.) From one year to the next, team offense has shown a correlation of .41. Teams' starting pitchers -- as measured by sOPS+ for each team's starters -- have shown a year-to-year correlation of .55.
Bullpen quality is much more fluid. The correlation from year to year, by bullpen sOPS+, has been just .30. I'm not sure that's the best stat to measure bullpen quality, given that it includes the mop-up men working meaningless innings. So if we look at the year-to-year correlations by bullpen Win Probability Added -- which puts much more emphasis on the key pitchers who are called in when the game is on the line -- the correlation drops to .20.
A correlation of .3 or .2 is certainly not randomness -- the past is some guide to future -- but it suggests that a team's architects can do everything right and still end up with a dud or do nothing at all and end up with an ace staff.
Consider, once again, the Cleveland bullpen. The club started with six returning relievers, all of them with recent history of fabulous success and none of them notably old. (Miller and Otero are the oldest at 33.) What happened to this half-dozen?
Cody Allen: The difference between last year and this year is about three runs. He has basically walked one extra batter and allowed one extra home run, which, by strict accounting purposes, has turned him from one of the game's best closers to a below-average one.
Andrew Miller: He was awesome all April -- a 0.00 ERA, and batters hit .162/.279/.216 against him -- before he went on the 10-day disabled list with a hamstring strain. When he came back, he was wild, and over the course of six appearances -- 25 batters faced -- he was hit for a .368/.520/.842 line. Then he went back on the disabled list with knee inflammation.
Tyler Olson: A lefty whose job is to get lefties, he unexpectedly dominated lefties and righties last year and didn't allow a run in 30 appearances. He has again dominated lefties this year! Meanwhile, righties are hitting .344/.432/.656 against him.
Dan Otero: A veteran sinkerballer who never strikes many guys out, Otero's walk and strikeout rates this year are consistent with his career averages. But he has allowed as many extra-base hits as he did all of last season, and he has been hit especially hard with runners in scoring position. He has allowed 33 baserunners, and 18 of them have scored.
Nick Goody: He was pitching great until he abruptly got hit hard in four straight games, after which he went on the disabled list. Now he's on the 60-day DL. There's no timetable for his return.
Zach McAllister: There's a pretty good case to be made that he's throwing well: He's throwing harder than last year, throwing way more strikes than last year and throwing more "edge" strikes on the black of the strike zone. He added a sinker and improved his ground ball rate. But there's also this: He has already allowed six homers (after allowing eight all of last season), and batters are hitting .395/.415/.816 with men on base, more than double the OPS he has allowed with the bases empty.
We started with one story -- Cleveland has a hot-mess bullpen, and that's why it is no longer an elite team -- but ended up with a bunch. There are a couple of guys who look worse only because of, literally, one or two balls that landed a few feet over instead of a few feet shy of the wall. There's a serious injury, and there's a less serious injury, each of which has disturbed Cleveland's plans in its own way. There's one pitcher, McAllister, who looks like a really strong bet to dominate in the second half. And there's one, Goody, who probably wasn't as good as we might have believed in the first place. All in all, it's ... still a pretty good bullpen.
Or maybe it's a bad one. Cleveland has as many walk-off losses this season as the team had all of last season. The Indians have begun the seventh inning tied 10 times, and they're 1-9 in those games. Last year, when they were trailing after five innings, they went 10-37. This year, they're 1-19.
It's never easy to tell until they make it easy on us all and actually play. It's a cruel joke to play on the team's front office, which has to figure out in advance which half-dozen coin flips to bet on. But there's something fun and pure about it, too: It's a part of the game that exists mostly outside of the front office's control. The onus is on the players, rather than some romantic vision we might have of omniscient general managers using their terabytes of data to make baseball predictable. The team that's good is the one that does better -- simple, nerve-racking, surprising and, occasionally, devastating.