Sometimes, predictive statistics tell you exactly what you would have expected. And sometimes, they tell you that a bullpen that was one of MLB’s worst last year is likely to suddenly become one of its best.

Like you, I haven’t forgotten the feelings of anger and helplessness I experienced while watching a parade of both terrible and underperforming relief pitchers blow multiple-run leads game after game during the regular season in 2018. But if you plan to continue to hear me out, I’m asking you to do me a favor and put those emotions aside for the sake of this article and focus on one important, mathematical fact: skills-based statistics are far more predictive than actual performance outcomes.

Suppose you roll a normal six-sided die three times, and all three times it lands on a six. The next time you roll that die, which is the more likely outcome: that you’ll roll another six, or that you’ll roll a number one through five? Those of you who understand probability would place your bets on the latter. More importantly, the fact that you rolled three sixes in a row wouldn’t have any impact on your prediction for the next roll. Keep this logic in mind; we’ll come back to it later.

This is the part where I let you be angry again by reminding you just how bad the Tribe bullpen was last season. Out of 30 MLB teams, the 2018 unit finished 25th in ERA (4.60) and 26th in FIP (4.53). The month of May was particularly bad, when Indians relievers allowed a whopping 65 earned runs, second only to the 70 earned runs allowed by a Marlins’ bullpen that threw 22 more innings during that span. On the whole, Cleveland cycled through 21 relievers in a desperate attempt to find stability that never truly emerged.

That said, there are a number of things we need to consider in order to try and predict how the relief corps will perform this year. First and foremost, we can’t factor in any relievers who are no longer on the team. That means forgetting about contributions from Cody Allen, Andrew Miller, Zach McAllister, Josh Tomlin and a host of relievers that contributed just a handful of appearances prior to being DFA’d during the season.

After that, we need to be cognizant of the fact that small sample sizes of data are neither reliable nor predictive. Anything less than a ten-inning sample size isn’t significant enough to give a clear picture of a reliever’s skill set, and even data samples two or three times that size are somewhat shaky.

For the purposes of predicting the strength of the 2019 bullpen, then, I’m only looking at the relievers who A) threw at least 10 innings during the 2018 campaign, and B) are still within the Indians organization. Conveniently, that set of parameters creates a seven-player list that lines up almost exactly with the list of those who find themselves atop the Tribe’s depth chart at Roster Resource. In order to form a more complete picture of the club’s projected opening day bullpen, I’ve added the recently-signed Alex Wilson to that list. Here it is in full:

Brad Hand
Dan Otero
Oliver Perez
Tyler Olson
Neil Ramirez
Nick Goody
Adam Cimber
Wilson

By all indications, this group of relievers was solid last year on the whole. Their collective ERA checks in at 3.73, which would have ranked 10th among MLB ‘pens last season. Their collective 3.88 FIP, meanwhile, would have tied the Dodgers for 8th place. Fangraphs’ FIP-based WAR calculation estimates that, had the Tribe trotted out these eight relievers all year long supplemented by only replacement-level arms, they’d have won about two and a half more games.

This already improves the outlook for the maligned Cleveland bullpen (and pretty dramatically so), but there’s one important notion we have yet to consider, and it’s by all means a game-changer. See, there’s one luck-driven statistic that FIP entirely ignores, and it just so happens to be one that points to an unsustainable amount of misfortune for the Tribe’s projected relief corps.

There’s one thing in particular that the aforementioned eight relief pitchers did really well last year on the whole: they managed to keep the ball on the ground. Were we to rank their combined 48.7% ground ball rate amongst bullpens in MLB last year, they would have been bested only by the Diamondbacks and their 50.6% figure. Considering the fly ball revolution that continues to drive a home run spike in today’s baseball game, that’s a fantastic skill for pitchers to possess.

Naturally, then, we’d expect that this relief group allowed very few homers in comparison to the rest of the league. And yet, to the contrary, they allowed a whopping 1.23 home runs per nine innings. That figure is nearly 14% higher than the league average mark of 1.08, and would’ve ranked as the 25th-best among major league bullpens in 2018.

Obviously, the anomaly lies in the fact that an unusually high portion of the fly balls hit off these eight pitchers cleared the outfield fence. Data show that the Tribe’s bullpen-to-be owned a monstrous 16.6% homer-to-fly-ball rate last season. For a frame of reference, consider that no bullpen in baseball outside of Cleveland posted a HR/FB% north of 14.5, and that the league average mark among relievers was about 12.1.

At this point, you might be wondering whether this particular collection of pitchers are just easier to hit with authority, making them prone to longer fly balls than others in the game. They aren’t. Data at Fangraphs says this group gave up a collective 34.1% hard contact rate last season, which actually checked in a few ticks below league average. In fact, the sabermetric community posits that homer-to-fly-ball rate is largely out of a pitcher’s control. Take a look at this excerpt from a post by Steve Slowinski of Fangraphs.com:

“[P]itchers generally don’t have much control over how often fly balls leave the ballpark. They play a major role in allowing those fly balls to begin with, but the difference between a ball clearing the fence and dying on the warning track is largely out of their control and it generally takes several hundred fly balls allowed for that luck to balance out…This isn’t to say that HR/FB% isn’t a skill, but rather that the true gap is much smaller than the observed values you might see over the course of a single season worth of data.”

Even if you don’t want to take Steve’s word for it, consider that the exact same group of relievers we’ve been examining (minus Cimber, who didn’t make his MLB debut until last season) posted a combined 12.5% HR/FB in 2017 across a similar number of innings. Factor in that the league average mark was actually two thirds of a percent higher that year. Now ask yourself which of these two scenarios is more likely:

  1. Some combination of a) the adjustments made by the league against these specific seven relievers and b) the declining fly ball distance-control skills of said relievers precipitated an environment in which hitters matched up against them were able to get the extra fly ball distance needed to turn nearly 33% more of those fly balls into home runs.
  2. There is significant random variance in HR/FB%.

If, like me, you buy into scenario number two, then you’ll probably also be willing to subscribe to the notion that HR/FB% isn’t a very predictive statistic. Fortunately, there’s a statistic that calculates FIP within a parallel universe where every pitcher owns a league-average HR/FB%. As many of you know, it’s called xFIP, and it’s a far more reliable indicator of a pitcher’s actual skill set than regular FIP.

As you might have guessed, the case for the Indians’ bullpen as one of the top five in baseball right now is based largely on the fact that the eight relievers I’ve been relentlessly analyzing, the ones projected to comprise the Tribe’s Opening Day relief corps, posted a sterling xFIP last season. To be precise, they combined for a 3.47 mark across 373 innings during the 2018 campaign; that would have tied the Brewers for third in all of baseball last season.

Of course, the Tribe ‘pen needed to cover more than just 373 innings last season. But even if the aforementioned group could deliver that same number of innings with the exact same set of skills, the Indians could run replacement-level relievers out on the mound for another 90 innings (in order to replicate the 463 relief innings the bullpen tossed last season) and still achieve a 3.66 xFIP, assuming a replacement-level skill set would yield the same 4.47 xFIP it did last season. Even that 3.66 mark would have eclipsed the Dodgers for fifth-best in MLB last season.

So don’t panic over the state of the bullpen. While the dice roll yielded an unquestionably pathetic outcome last season, consider the actual probability looking forward than the most recent outcome looking back. Predictive stats tell us that the Tribe has the makings of a relief unit that’s not just good, but actually elite.