The extent to which the start of the Atlanta Braves’ 2022 season has mirrored the start of their 2021 season has been somewhat uncanny, and also very unpleasant. Thirty-one games in, the 2021 Braves were 15-16. Their subpar performance to date had resulted in them shedding about four wins off their preseason expectation, and about 20 percent in playoff odds. 31 games in, the 2022 Braves are 14-17. Their subpart performance to date has resulted in them shedding about four wins off their preseason expectation, and about 20 percent in playoff odds.
The above is a little misleading, because the 2021 Braves came into the season with lower expectations and lower playoff odds — their 15-16 start took them to 83 projected wins and a depressing playoff odds mark of 38 percent. The 2022 Braves, meanwhile, have somewhat more rope due to both their roster and the expanded playoff situation — their 14-17 start still entails 88 projected wins and playoff odds of 69.9 percent (nice, but also not nice given where they started). I think it’s kind of a similar bummer relative to preseason expectations, though you’d prefer where the 2022 Braves are, one game worse and all, to where the 2021 Braves were at the same point.
It’s not just the results that have been similar — the “how” is pretty much the same as well. Through their first 31 games, the 2021 Braves had the league’s best xwOBA, but only its seventh-best wOBA, entailing the second-biggest xwOBA underperformance among MLB teams. Through their first 31 games, the 2022 Braves have the league’s fourth-best xwOBA, but are tied for its 11th-best wOBA, entailing the fourth-biggest underperformance among teams. While not everything is the same (for example, the 2021 Braves had way better defensive performance through 31 games, but the 2022 Braves have much better pitching), there are some unfortunate throughlines:
- 2021, through 31 games: 1st in xwOBA, 7th in wOBA, 12th in RE24, 16th in WPA
- 2022, through 31 games: 4th in xwOBA, T-11th in wOBA, 14th in RE24, 22nd in WPA
The chain of four stats above goes from inputs (what hitters have the most control over) all the way to context-specific results (what hitters definitely do not have any control over). You’re likely familiar with xwOBA (an assessment of what a hitter’s wOBA should be based on the exit velocity and launch angle of their batted balls, plus walks, strikeouts, and HBPs) and wOBA (the actual weighted results of what the hitter has received, placed on the OBP scale). RE24 is “the change in run expectancy associated with the hitter” — essentially every base-out state (there are 24 of them) has a default run expectancy, and after the hitter’s PA, runs do/don’t score, and run expectancy increases/decreases; a hitter’s RE24 is just the sum of those run expectancy changes from the hitter’s PAs. A bases-loaded single with two outs is worth way more RE24 than a bases-empty single with two outs, whereas in wOBA they’re both just singles. WPA takes RE24 a step further, looking not just at run expectancy, but win expectancy, which factors in the base-out state but also the score and the inning. A bases-loaded two-out single when up by 10 adds essentially no WPA; a bases-loaded two-out single that yields a walkoff win yields a boatload of WPA. This refresher aside, the point is that early in both 2021 and 2022, the Braves had...
...fantastic offensive inputs, which became...
...good-but-not-great offensive outputs, which gave them...
...mediocre (not even good) impacts to run expectancy, which came...
...in an array of situations such that the Braves have benefited from less than the “default” amount of win expectancy from their offensive inputs or outputs or even the base-out states in which those outputs have occurred.
“Tailspin” is probably too harsh a word to describe a 14-17 record throughout which the Braves have never lost more than two games in a row (they also haven’t won more than two games in a row), but it is probably not too harsh a word to describe how Marcell Ozuna has been smack dab in the middle of all of the above.
Want to talk about uncanny similarities? Check this out:
ISO within .001. xwOBA within .004. wRC+ within 3. Same costly, negative fWAR (albeit in way fewer PAs in 2022).
Oh, but that’s not all. Let’s do those same 2021 versus 2022 bullets again:
- 2021 Ozuna, through May 25 (the day of his hand injury): 72nd percentile xwOBA, 32nd percentile wOBA, 29th percentile RE24, 6th percentile WPA
- 2022 Ozuna, so far: 63rd percentile xwOBA, 32nd percentile wOBA, 1st percentile RE24, literally the worst WPA in baseball.
Okay, I lied. It’s not similar. It’s worse. Ozuna has taken slightly worse relative inputs, gotten the same horrid outputs on them, and somehow... has outperformed literally every other batter in MLB in 2022 so far at robbing win expectancy from his team.
I could go into a whole side-track here about how unlikely this is, but given that the season isn’t in the books yet (and thank your deity of choice for that), it would feel a little weird, given that circumstance will probably reverse itself, at least somewhat. Still, even so far in 2022, we know the following:
- The relationship between xwOBA and wOBA, even through just about six weeks of the season, has an R-squared of around 0.6, which means that about 60 percent of the variation in context-neutral outcomes (wOBA) is explained by batter inputs (xwOBA). Ozuna hasn’t been killed by an xwOBA-wOBA gap as much as around 50 other players this year, but he’s still in the 15 percent of players most affected by this particular way to have results fail to match inputs so far.
- If we convert RE24 to a per-PA measure (since it is a counting stat), the R-squared between xwOBA and RE24 is around 0.47 so far in 2022. If we use wOBA instead of xwOBA, it goes up to 0.74, which makes sense. Basically, inputs are about half of RE24, outputs that differ from inputs are another fourth, and the rest is sequencing. (This is horribly non-scientific and not actually correct, since I didn’t run this in a multivariate model, but it’s at least a useful heuristic? Maybe?)
- The relationship between RE24 and WPA has an r-squared of around 0.73 — again, around three fourths of WPA accumulation so far in 2022 is independent of inning/score. If you skip past RE24, then the r-squareds for xwOBA to WPA and wOBA to WPA are 0.31 and 0.52, respectively. The fact that variance in offensive input quality explains only 31 percent of actual win expectancy accrued by hitters through six-ish weeks of the calendar is why baseball is A) fun; B) frustrating; and C) reliant on much larger samples than 30ish games to draw inferences about team quality when looking at things like the standings.
Anyway, back to Ozuna. We know that the relationship between xwOBA and RE24 is highly imperfect, but if you build a model that infers a player’s RE24 based on his xwOBA, then Ozuna’s -11.56 RE24 should be +1.23. This difference of nearly 13 runs of RE24 is egregious, not only the biggest difference in MLB, but bigger than the next-worst RE24-based-on-xwOBA underperformer by over 40 percent. If you instead build a model that infers a player’s RE24 based on his wOBA, it still looks just as bad for Ozuna: his -11.56 RE24 should be -3.26 by this model, and that difference of over eight runs is nearly 20 percent higher than the next-worst wOBA-to-RE24 underperformance.
Given that he’s the worst at WPA so far, these things cascade into some egregious underperformance there, too. If you use xwOBA to infer WPA, Ozuna should have a neutral-ish 0.06 WPA instead of -1.35, once again, the biggest difference in MLB. If you use wOBA to infer WPA, he at least gets some reprieve as Aaron Hicks has a bigger gap... but he’s still underperforming the should-be-WPA by 1.000, which is about two wins’ worth of WPA. (Each team earns net .500 WPA in a win, taking their pre-game 50 percent chance of winning and turning it into a 100 percent chance.) If you use RE24 to infer wOBA, Ozuna isn’t horrible, since both his WPA and RE24 are bad, yet he’s still underperforming his RE24-based WPA by over 0.2, or about half a win.
So, there you have it. Ozuna has been a calamity dropped on the Braves’ win expectancy, and it’s not even really his fault.
But I suppose it’s not entirely not his fault, either. This isn’t the article to discuss everything that’s changed for the worse for him so far in 2022, but Ozuna’s calamitous lack of WPA is coinciding with a change in approach that’s getting him away from the Braves’ general offensive strategy of swinging only to do maximum damage. While his hard-hit rate is fine, his average exit velocity is down, and his xwOBACON has really dipped (still well above-average) as too much of his contact is angled slightly downward: his grounder rate is above 42 percent after sitting around 37 percent in 2020-2021.
That’s what happens when he does make contact, but it’s the result of some poor decisions: Ozuna currently holds his highest chase rate, as well as his highest o-contact and z-contact rates. In theory, neither o-contact nor z-contact have to be bad, but this clearly isn’t working, as the extra contact he’s making is of low quality. You can see this at an obvious surface level: his walk rate has plummeted to a 58 BB%+ (i.e., nearly half of league average), his lowest since his 2013 rookie season, while his strikeout rate has fallen to a 76 K%+, his lowest in a season. You can actually very easily see the reason for the worse, grounder contact — the left side are Ozuna’s swings is 2021 (which, for the season, looked very similar to 2022, but with different decisions), normalized to the same number of swings as he’s taken in 2022, while the right side is 2022. There are more swings at stuff that isn’t even that borderline in 2022 away and down, and the overall swing area isn’t as focused to down-the-middle stuff as before. For another comparison, look at Ozuna’s swing pattern in 2020, at the bottom. Sure, there’s some chasing, but for the most part, Ozuna picked things he could mash out of the park... and then mashed them out of the park. It’s harder to do that these days, but it’s still better than the alternative.
Bad swing decisions aside, though, none of it explains the absurdity of his wOBA-xwOBA gap, and his wOBA/xwOBA-RE24 and wOBA/xwOBA-WPA gaps. That’s different. It might happen to anyone, though it probably won’t. But it is happening to Ozuna, and here we and the Braves are.
This article was developed on Wednesday. After most of the data were pulled together, there were two small developments. First, Braves head honcho Alex Anthopoulos offered this tidbit in an interview with The Athletic:
His exit velocity is still really good. The biggest thing is his walk rate is down. He’s putting balls in play. If you’re swinging at a zone rather than swing-and-miss, you’re still alive in the at-bat. Slower start, obviously — him and some other guys. But he’s still hitting the ball hard. His average is down, but that’s because if you’re swinging a bit more outside the zone, it’s obviously not going to come off as well.
So, everyone knows what’s up. That’s good. The first step to fixing a problem is identifying it. (The next step is acknowledging that it’s not just the chase rate, it’s the decisions leading to it — why swing if you can’t kill it?)
The second was that, for the first time this season, Ozuna was dropped down in the lineup (he had hit third in one game, but otherwise spent the year at cleanup). Did it do anything? Well, no. It was more of the same. Ozuna finished Wednesday’s game with, once again, the lowest WPA among his teammates. He went 0-for-4 with four balls in play. In his highest-leverage PA of the day, something that could have given him a chance to reverse all the crap I wrote about above, he barreled a ball: 103 mph, 33 degree launch angle, traveling 379 feet. It was caught. Ozuna ended his day with a .000 wOBA and a .514 xwOBA. He now has the worst WPA in MLB by nearly .030, which is the difference between the second-worst WPA and the 17th-worst WPA. His xwOBA underperformance is now essentially in the top (bottom?) decile of underperformers this season, and none of the batters underperforming worse than him have as many PAs as he does.
So it goes.