clock menu more-arrow no yes mobile

Filed under:

Too in depth on depth: the Braves and 2023 LF/DH options

The Braves’ LF situation is a mess. Should they have done something different to bolster the position?

MLB: JAN 21 Braves - Braves Fest Photo by David J. Griffin/Icon Sportswire via Getty Images

Friends, I must inform you that there has been a disagreement. Now, I know what you’re thinking. A disagreement? In this blog’s comment section? No way, that never happens. But yes, here we are. Not long ago, in the comments section of a post containing the results of a poll regarding whether the Braves “did enough” this offseason, battle lines (of no consequence) were drawn. On one side, those arguing that the Braves didn’t do enough because of a lack of suitable options for left field and DH; on the other side, those who believed this wasn’t a failing. I, myself, align to the latter — in my view, albeit one reflecting a prior rather than something rigorously tested before I came to that conclusion, was that the difference between the left field options the Braves did acquire, and do have in house, versus the options they could have reasonably obtained but chose not to, was inconsequential.

But just having an untested prior is no way to go through life. So, in something that took up probably way too much time I should’ve spent on other things in my life, I present to you this post, where I ran a bunch of numbers to see whether I was right, and the Braves likely couldn’t have reasonably improved their LF/DH outlook... or whether I was wrong, and there was an obvious path to meaningful improvement of their current options.

The Cast of Characters

Let’s start at the beginning. The Braves came into the offseason with the known LF/DH options of Marcell Ozuna and Eddie Rosario. The less said about these guys’ 2022 seasons, the better — but both were still under contract for 2023. The Braves supplemented this crew with a few also-not-starter-quality options: they acquired Sam Hilliard from the Rockies in November, they signed Jordan Luplow to a $1.4 million contract in December, they traded for Eli White in December, and then they signed Kevin Pillar to a minor league deal in January. At the same time, there were a bunch of guys the Braves didn’t sign for more or less the same outlay. Of the names floated in this case are guys like Ben Gamel (Rays minor league deal, February), Tyler Naquin (Brewers minor league deal, February), Franmil Reyes (Royals minor league deal, February), Jason Martin (NC Dinos of the KBO, December), and Donovan Solano (Twins, major league deal, February), among many others.

As part of said disagreement, I proposed the exercise I’m going to describe in this post. The exercise consists of comparing the guys the Braves do have, to a subset of the guys they didn’t get. Specifically, for the purposes of this exercise, we’re comparing:

  • What the Braves did: Ozuna and Rosario, followed by some combination of Hilliard, Luplow, and Pillar.
  • What the Braves didn’t do: Ozuna and Rosario, followed by some combination of Martin, Reyes, and Solano (as chosen randomly by EccentricATLFan).

(Note: I initially offered to run a version with three backups/depth options, as well as four, which would’ve added Eli White to the Braves’ slate and Ben Gamel to the other scenario. However, I didn’t end up running this alternative, for reasons that will hopefully be clear with the below, though it’s still possible to do so if someone still thinks given the below that this will somehow make a difference.)

Let’s summarize these eight guys, just so we have a starting point.

In case you don’t know much about why these guys are expected to do what they are, here’s a brief overview:

  • Ozuna has been awful for two years, but his quality of contact and general track record suggest he probably won’t be as awful again. He doesn’t give you anything defensively.
  • Rosario had a nightmare 2022, and a lot of his recent track record is more akin to a bench piece, which is why his projections are pretty terrible. Being non-starter quality and then having an awful year is a great way to fall out of the league.
  • Hilliard has never really hit despite some interesting offensive qualities, but has enough speed-and-defense to be a decent backup.
  • Luplow is less of a defensive backup, but has hit at a league-average rate in his career and has had both awesome defensive and awesome offensive partial seasons, once upon a time but not all that recently.
  • Pillar is kind of like a more veteran Hilliard, but older and therefore probably a slightly better non-impact bat to go with not-quite-as-speedy-or-defensive qualities.
  • Martin hasn’t played much in the majors, but given his Triple-A stint last year, seems like a low-OBP, fine enough defensive outfielder.
  • Reyes is probably limited to just DH, but had an okay enough bat until 2022.
  • Solano has had a long and varied career and has transitioned from a no-bat defensive whiz to a quality bat that’s a lot less interesting defensively. He has barely played the outfield in his career.

In aggregate, the projection systems (Steamer, ZiPS, and a blend) have all of these guys as somewhere between decent bench guys and low-end regulars; IWAG is less sanguine about most of them, collectively seeing them as bench guys or perhaps not even worth rostering. The biggest disputes are about Ozuna and Luplow, but you can see the table above as well as I can. Anyway, because we want to compare the second group in that table to third group, the key point is that no projection system sees them as being particularly different. That’s actually kind of the answer — you could stop right there, if you wanted! Even without any playing time considerations, extra modeling, etc., the Hilliard/Luplow/Pillar group averages 1.3 WAR/600 across players and systems, while the Martin/Reyes/Solano group averages... 1.3 WAR/600 across players and systems. Donezo.

The method

But we’re not donezo. Not even close. We’re going to overkill this question. Here’s what I did.

I set up a mini-model with IWAG to do the following.

  • First, pick one of Ozuna or Rosario. That guy gets the first 100 PAs. If he puts up more than 0.33 WAR (aka, 2 WAR/600) in those PAs, as determined by IWAG’s probability distribution of player production for 2023, he continues onward at that same rate, until IWAG says he hits his PA limit for the year (rounded to the nearest 100, for simplicity). To be clear, the PA limit is essentially a measure of injury propensity — IWAG estimates how many PAs the player can get in any given run, based on the player’s injury history.
  • Then, if that guy either doesn’t clear the 0.33 WAR/100 threshold, or hits his PA limit, pick the other of Ozuna/Rosario, and do the same thing.
  • Once Ozuna and Rosario have both exhausted their PAs, pick a guy at random from the set of Hilliard/Luplow/Pillar or Martin/Reyes/Solano. Do the same thing for them, switching to the next randomly-selected guy.
  • Do this until a run hits 1,000 total PAs. Why 1,000, and not 700, which is the standard central estimate for PAs available at a given position? Because we’re doing both LF and some part of DH here, and the Fangraphs Depth Charts currently estimate roughly 300 DH PAs for the Braves not taken up by Travis d’Arnaud or Sean Murphy when they’re not catching. Note that even though I am shunting about 30 percent of the estimated PAs here into DH, I am not applying a corresponding fielding value penalty nor any kind of hitting penalty for those PAs, because the point is just to compare scenarios. (If you’re wondering about a hitting penalty, there’s some evidence that playing DH reduces a guy’s hitting output relative to not playing DH.)

The results

I more or less spoiled this already, but yeah, it doesn’t matter. In the “Guys the Braves have in hand” scenario, the mean WAR/600 (across 1,000 PAs) is 2.0, the median is 1.9, and the mode value, rounded to the nearest 0.5 WAR, is 1.0. (Why is the mode so different? You’ll see in the density plot a bit below that most results are in the 1ish range, but there aren’t many worse results, so the skew of the data is entirely towards higher WAR values.)

Meanwhile, in the “Other three guys the Braves could have signed” scenario, the mean WAR/600 is 1.8, the median is 1.7, and the mode is 1.5. This scenario was more concentrated in midrange outcomes, but with fewer top-end outcomes, which is mostly driven by the fact that the IWAG distributions see Luplow as having the best chance of all of these six depth pieces of racking up a lot of value in short stints.

This density plot illustrates the differences between the two scenarios, such as they are, pretty well. I smoothed the visual less than I normally do for projections just so it wasn’t quite two slightly different rainbow-shaped arcs, but you get the idea.

Blue line = scenario with the guys the Braves actually signed; red line = scenario with guys the Braves didn’t sign.

But wait, there’s more. If you were looking at the paragraph above, and perusing the Braves’ projections from any point this offseason, something might be bugging you. Specifically: how is it that the Braves, who are projected for only 1ish WAR per 700 PAs at both LF and DH, getting means and medians of 2 WAR per 600 in this exercise... especially when IWAG is more down on Rosario, Ozuna, and some other guys relative to Steamer and ZiPS?

It’s a great question, and I have a solid answer for you. Here it is: forcing the team to rotate players every 100 PAs if they don’t play well is a huge boost to the WAR they can achieve. Why? Because, well, few of these guys have large portions of their production distributions for 2023 that amount to disasters. Rosario does, while Reyes and Ozuna both have problems with their production if they don’t hit because they also don’t field much. For most of the other guys, their legs, gloves, and arms give them a non-awful floor. By using a comically-large hook and removing a guy from playing time if he’s below average, the assumptions described above basically churn through all five options super-fast, until they hit on a guy that is going to produce. Under the IWAG probability distributions, here are the chances that a guy produces at a rate that has him stick past his first 100 PAs:

  • Ozuna: 34 percent
  • Rosario: 25 percent
  • Hilliard: 16 percent
  • Luplow: 13 percent
  • Pillar: 14 percent
  • Martin: 62 percent (but should be noted, these are mostly barely above the threshold, so Martin offers a projected steadiness with little upside beyond 2 WAR/600)
  • Reyes: 3 percent
  • Solano: 57 percent (with a lot more upside than Martin)

In the first scenario with Hilliard/Luplow/Pillar, there is a greater chance that someone “breaks out” when you churn through all five guys than that they play to their central estimates or worse. In the second scenario, given the percentages with Martin and Solano, it’s very likely that those guys will give you okay production.

But here’s the thing: this sort of hot-swapping after 100 PAs isn’t that likely. In reality, where we don’t have a model that tells us a guy’s production level now and forevermore, 100 PAs probably isn’t enough to make a useful judgment unless teams have access to specific internal data that gives much, much better immediate-term projections than what we have in the public sphere. (The stuff they have is probably somewhat better, but teams give enough PAs to crappy players who produce crappy WAR values that it’s hard to believe they’ve got this sort of near-term forecasting on lock.)

So, I changed it up. I made the limit before swapping 200 PAs, which is something I would recommend given my understanding of the extant baseball research. Lo and behold, this fixes both the problem with the prior scenario, and still doesn’t show much difference between the two scenarios.

  • Guys the Braves have in hand: mean = 1.2 WAR/600, median = 0.8 WAR/600, mode (rounded to nearest 0.5 WAR) = 0.5 WAR/600
  • Guys the Braves didn’t sign: 1.4 WAR/600, median = 1.4 WAR/600, mode (rounded to nearest 0.5 WAR) = 0.5 WAR/600 (though really 1.5 WAR/600 is pretty close)

I suppose you could argue that the Braves should’ve gotten a “steadier” option like Solano, or someone else with a beefier range of projected outcomes around 2 WAR. But they arguably seemed to have preferred Luplow’s short-stint upside instead. The weighted average production isn’t very different, not meaningfully so.

Blue line = scenario with the guys the Braves actually signed; red line = scenario with the Braves didn’t sign.

Anyway, I don’t know if this has changed anyone’s mind. Even if it didn’t, hopefully it gave you some food for thought. Perhaps one reason why the differences aren’t dramatic, or perhaps barely present: Ozuna and Rosario are still going to play and get first dibs, presumably. That seriously limits the potential for depth to matter, even if we assume there are 1,000 PAs to go around, and both of those guys might lose their starting gigs after 100-200 PAs each. After all, if either one of Ozuna or Rosario produces at something resembling a reasonable rate, chances are the depth won’t get a chance to play all that much. If both do, the depth is barely going to get used until someone gets hurt. (And someone will get hurt.)

Given the above, I think the only thing to hang one’s hat on is that the Braves didn’t do enough because they didn’t get any likely-to-be-average options for LF. Beyond that one sticking point, I’m not sure there’s much of a difference between the guys they got and the guys they didn’t.

Areas for further analysis could be doing a four-player variant (e.g., adding Eli White to one scenario and Ben Gamel to the other), though I’m not sure it’d change anything, or adjusting the point estimates from IWAG to better reflect Fangraphs Depth Charts... although again, this won’t have much of a comparative impact — it will likely make any differences between the two scenarios even more minimal, given that higher projections mean less churning through guys, and therefore less reliance on depth.

Sign up for the newsletter Sign up for the Battery Power Daily Roundup newsletter!

A daily roundup of Atlanta Braves news from Battery Power