Were Relievers Worth Their Beans In 2020?

Last offseason, the Braves splurged on their bullpen, spending a combined $57.75 million on Chris Martin, Will Smith, Darren O’Day and Josh Tomlin. Coming into the 2020 season, their bullpen represented approximately a third of the team’s overall salary commitments.

Meanwhile, the Kansas City Royals found themselves on the other end of the bullpen spectrum; a 103 loss season didn’t warrant any huge expenditures on their relief corps. Instead, the Royals bought low on a couple of hurlers that had fallen on hard times: Greg Holland and Trevor Rosenthal. The duo, rather unexpectedly, helped anchor a bullpen that posted one of the Majors’ lowest ERAs.

On the East Coast, the Orioles loaded up their pen with a ton of unheralded arms. Props to you if you could name, off the top of your head, the pair of Orioles’ relievers that finished the season with the second-most appearances on the team (Paul Fry and Travis Lankins. Sr with 22 apiece). Though the Orioles’ pen featured many names even the biggest baseball fanatic wouldn’t know, the unit still combined for a 3.90 ERA, ninth best in the Majors.

There’s a question that pops up comparing these three strategies: What’s the best way to create a bullpen?

Is signing a bunch of high-end arms the right move or is it better to ink reclamation projects and hope for the best?

Well, it depends.

Relievers are a fickle bunch, one year, they can dominate, the next, fall off the face of the Earth. What worked in Season 1 is far from a guarantee to work in Season 2. It’s in their nature to be volatile.

Their success is contingent on the team that signs them. Some clubs opt to provide data-driven solutions to players while others believe in his natural abilities. From what I’ve seen, the former tends to enjoy more success than the latter.

Look at how this year’s World Series clubs constructed their relief corps. The data-heavy Dodgers brought aboard Jake McGee, Adam Kolarek, and Dylan Floro, all of whom helped the franchise win its first World Series since 1988.

Meanwhile, the Rays collected castoffs from virtually every organization and transformed them into high-impact arms. John Curtiss, Aaron Slegers, Ryan Thompson, the list goes on.

Tampa Bay bought low on cost-controlled reclamation projects and connected them to the game’s best player development program in order to maximize their effectiveness. But, is that really the best way to build a bullpen?

There’s a couple of factors one considers making this decision.

  1. Is a team looking to contend for a postseason berth? If so, a team’s front-office might opt for the safer option and venture into the free agent market. But buyer beware! If the 2020 Phillies’ bullpen has taught us anything, it’s that signing a bunch of arms isn’t the wisest move.
  2. Does a GM have faith in his analytics and player development departments? It’s essential that both departments are strong. Analytics, to identify a target and player development, to help a reliever reach his ceiling.

But, is what the Rays did smart? Well, probably, because they’re the damn Rays. Any team that hires one of Fangraphs’ senior analysts and plays a style of baseball so analytically-savvy that it implores boomers to write hit pieces are definitely better off for it.

I wanted to take a look around the league and check if the Rays’ model was a good idea.

I was curious to see if there’s any correlation between a team’s top relievers’ salary and their production. To do so, I focused my analysis on the relievers that finished top-five on the team in games. I dubbed this group “the core”, as in the team’s core group of relievers. I also removed any “openers” from this group as that could have added some unnecessary noise into this study.

Considering the variability on how many relievers each team uses per season, five seemed like an acceptable benchmark.

To cement my “core” qualification, Baseball Reference designates five players as the core of a team’s bullpen and so, who am I to argue with Sean Forman?

I then compared a player’s non-prorated salary to his 2020 metrics and ran a correlation matrix.

Below is each team’s core’s bullpen metrics.

And here’s my overall correlation matrix.

Here’s a link to a Google Sheet with all the relevant research. (Excuse Google Sheet’s poor formatting options)

I made a couple of conclusions based on my correlation matrix.

  • Relievers earning more performed substantially better in ERA, FIP, xFIP, average EV against, WPA, and saves.
  • The amount of games a reliever pitched in are inconsequential to his salary.

So, what to make of this data? From quick glance, it’s pretty evident that paying relievers high salaries makes sense.

In my correlation matrix, having a higher salary correlates well to having a better ERA, FIP, xFIP, average EV against, BB/9, WPA, and xFIP. Meanwhile, a player’s salary doesn’t correlate too well with Saves and Barrel %. Most of the aforementioned stats are important metrics that I use to judge a reliever’s performance.

Saves is the outlier. It’s certainly not a stat that I trust to give me an accurate representation on what happened on the field. I only included it to satisfy my curiosity: do teams throw their most expensive arm into the 9th inning?

It turns out that wasn’t the case. Over the last decade, teams have become more adverse to saving their best reliever for the final inning and the data shows that 2020 was no exception. Saves can tell you which pitcher tossed the final frame, but there’s another metric that I trust more to tell me an accurate picture.

gmLI gives us an idea at the pitcher’s leverage when he enters the game. It tells us how much a manager trust his reliever. Does he only run him out when the game’s on the line or only for mop-up duty? Does a manager have any bias towards choosing high-priced relievers or does he pick the best arm out there, regardless of price?

gmLI’s Correlation Coefficient with Salary was only -0.02, indicating that managers chose both high-priced relievers and low-priced relievers during a game’s most impactful moment.

Although this study was certainly helpful in determining the relationship between a player’s 2020 salary and their results, there’s a lot of noise that could certainly impact the final result. My job isn’t just to report the final conclusion, it is also my unfortunate duty to poke any possible holes into my delicately-crafted study.

While this analysis looks promising on the surface, it isn’t without its own warts. A 60-game season doesn’t show a player’s true performance, much less, a reliever’s. One bad game could’ve ballooned his end-of-the-season line.

Furthermore, the issue of arbitration also pops up. A reliever might be one of the most-expensive players on the team, but only because his team control is winding down. Arbitration handsomely rewards only the relievers with high saves/holds totals. That being said, due to the ways the arbitration suppresses many relievers’ salaries, I’m willing to designate these expensive relievers as mere outliers.

The Athletics started selling high on players that were rewarded for saves through arbitration in the early-2000s. When their “closers” were receiving higher paychecks through the arbitration systems, thus putting them out of Oakland’s budget, Oakland sold high on them, reaping young, cost-controlled talent. Billy Taylor, Jason Isringhausen, Billy Koch, and Kieth Foulke combined for 147 saves throughout their seven years in the Bay Area; all four were quickly shipped out at the apex of their value (Smart Baseball).

Despite all that being said, I feel relatively safe in saying that this study had its desired effect: to check if a reliever was worth his beans in 2020. And it looks like they were!

This conclusion matches up with my on-paper theory. Where would the Braves be without Mark Melancon anchoring their back-end? How would the White Sox have fared without Alex Colome and his 0.81 ERA?

That being said, there are plenty of horror stories about teams that have spent heavily on their bullpens, only to see them implode in the regular season (Looking at you, 2020 Phillies).

Instead of spending blindly on relievers, it’s essential that they are paired with quality analytics/PD departments that could maximize their value. Without data-driven departments implemented through the organization, the reliever has a high risk of floundering.

If only the Rockies took my advice.

In the 2017-2018 offseason, the Rockies landed a trio of proven relievers: Bryan Shaw, Jake McGee, and Wade Davis. Shaw and Davis were fresh off with impressive runs with AL Central teams while McGee had hurled 108.2 impressive innings with the Rays the two seasons beforehand. We all know what happened next.

Instead of blaming their demise on The Coors Curse™, the cause for their struggles falls on the organization’s failure to understand how each player’s arsenal would play at Coors Field. BaseballCloud’s own Wyatt Kleinberg took a look at this phenomenon a couple weeks ago.

Davis, Shaw, and McGee all rely on a high-spin fastballs. At a normal altitude, that spin rate bodes well for their performance. But, at the Mile High Stadium, they were clobbered. 108 million dollars quickly went down the drain.

With all their money devoted to their bullpen, the Rockies failed to augment the rest of their roster, a move that’s still having ripple effects several years later.

In the last couple offseasons, teams have won the offseason by stocking up on relievers, only to see them fail to produce in the regular season. The 2018 Mets. The 2020 Angels. Each of those franchises didn’t have an above-average analytics department to drive their free-agent spending and it showed in their record at the end of the season.

Paying relievers is a risky endeavor. Instead of blindly spending tens of millions on a player that might top out at 80 innings each season, the key to having a successful bullpen is to identify unheralded arms and have a team’s analytics department do the rest.

Opening Picture Credit: Scott Varley of MediaNews Group/Torrance/Daily Breeze

Leave a Reply

Powered by WordPress.com.

Up ↑

%d bloggers like this: