Regression Tried to Kill The Yankees. Is It About To Save Them?

Will regression to the mean, which has throttled DJ LeMahieu, kick start Aaron Judge? (Jim McIsaac / Getty Images)

We’ve watched the Yankees play 108 games this year, and we’ve seen them defy their pre-season projections time and time again. This team was supposed to lead the world in both position player WAR and pitching WAR. They are second in the league in pitching WAR, thanks to very strong performances across the board, but Fangraphs has them 17th in position player WAR, which is not nearly good enough for this team. What’s to blame? Plenty of things. Without the juiced ball, Gleyber Torres and DJ LeMahieu have become much less productive. Gio Urshela added a bunch of swing-and-miss to his game without adding any power. Clint Frazier traded slugging for contact. Aaron Hicks missed the majority of the season. Luke Voit missed the majority of the season. Brett Gardner has started more games in the outfield than anyone else. Essentially, this team’s position players have underachieved across the board.

I’ve said this plenty on our podcast, but thinking that this team is bound to start producing better on offense than they have over the course of this year does not make you crazy. They choose between a multiple-time All Star and the reigning AL home run king at first base. Their second baseman led the league in batting average, OBP, and OPS last year. Their shortstop went to multiple All Star games before turning 23. Their third baseman had a 133 OPS+ between 2019 and 2020. The outfield (when healthy and actually permitted to play a defensive position) contains two of the most recent players to have 50+ homer seasons, and one of the best lefty power bats in the game. What gives?

The best answer I can give you here is that the Yankees are victims of regression in some cases, and victims of a lack of regression in others.

Victims of Regression: Is This Who You Really Are?

DJ LeMahieu is the major victim of regression to the mean (which will come for us all sooner or later). In 2017 and 2018, the two seasons before DJ joined the Yankees, he posted a wRC+ of 91 and a wOBA of .333. That’s just below league average. This year, he’s posting a wOBA of .314 and a wRC+ of 99. He’s underperformed a bit compared to his xwOBA of .335, but he’s not in the same universe that he was in 2019 and 2020, when his wOBA was .388 and he was 46% better than the average hitter. This season should not come as a surprise, though. DJ has one season of 100 or more games outside of Coors Field in which he was better than league average offensively, and in that year he outperformed his xBA by 23 points, and his xwOBA by 9 points. Would expected DJ LeMahieu have been good in that year? Yes. He’d be better than 2021 LeMahieu. The key issue here is that DJ played 195 phenomenal games with the juiced ball and some great luck that completely skewed our expectations for him.

This graph, which compares DJ’s batting average on balls in play (BABIP) to his ground ball rate, goes a long way toward explaining why DJ’s regression is not particularly surprising. Generally, this graph shows an inverse relationship between the likelihood that a batted ball is a hit (BABIP) and the likelihood that said ball is a ground ball (GB%). That’s not surprising. Defenses know exactly where to position themselves to field ground balls (see: reasons why banning the shift would ‘work’), and the density of defenders is much higher in the infield, where ground balls are.

For three consecutive seasons, DJ had rather weird batted ball luck relative to how often he was hitting ground balls. In 2018, he posted a career low ground ball rate, and a career low BABIP. That should not happen. He did make more soft contact in that year than in any other full season in his career, and posted a very low line drive rate, but his hard hit rate was fine, his slugging was present, and he was right in line with his xwOBA. Super weird.

In the following two years, the trend got weirder. His BABIP improved in spite of his ground ball rate increasing. That’s not supposed to happen. Sure he was hitting the ball a bit harder than he had in prior years, but no harder than he was in 2018 (on an average exit velocity basis). All those ground balls becoming hits, when they probably did not deserve to, did him a ton of favors.

2020 DJ LeMahieu is one of the luckiest baseball players on the planet. First, the season was only 60 games, so he hardly had time to regress to the mean. In 2020, DJ outperformed his xBA by 56 points, his xSLG by 137 points, and his xwOBA by 68 points. That’s basically impossible over 162 games. The key, of course, is that he did not have to play 162 games, so he looked like Ted Williams.

Let’s be clear about one thing. In 2019 and 2020, DJ LeMahieu had the peripherals of a very good hitter. If he had played to his expected stats, he still would have hit just over .300 and slugged just over .450. Those would both be slightly better than his career averages, I think the deadened ball and defensive alignments bear significant responsibility for DJ’s regression. He is actually elevating the ball more (by average launch angle, not ground ball rate) than he was in 2019, but he is hitting the ball hard 31.2% less often this year than he did in 2019. That means that ground balls don’t sneak through the right side or up the middle like they did in 2019 and 2020, and it means DJ turns into a true talent .270-.280 hitter. DJ’s expected stats this year emphasize that he’s not as bad as he has played, but his expected and realize stats both demonstrate that this year is much more the rule than the exception. Generally, guys don’t become brand new, MVP-level players in their age 30 seasons, and DJ LeMahieu, unfortunately, did not break that trend.

There’s more than regression to the mean going on with Gleyber Torres, but that’s definitely at play.

In 2018 and 2019, Gleyber posted an xwOBA of .334 each year, but posted realized wOBAs of .349 and .358. The kid was getting lucky. This extends to his xBA, which he outperformed by 19 and 25 points in 2018 and 2019 respectively, and his xSLG, which he outperformed by 28 and 55 points in those years, as well. Gleyber Torres took advantage of the rocket ball really nicely. In 2020, we started to see him come back down to earth. While he was still better than league average (106 wRC+), he lined up perfectly with his xBA of .243 and slightly underperformed his xSLG of .381, coming in at .368. The concerning aspect of all of this is that his xwOBA hardly changed. It was only 3 points lower in 2020 than it was in 2018-19. Gleyber just found his level.

Is Gleyber Torres a .246/.334/.349 (.683 OPS) hitter, as he’s been in the last couple of years? I doubt it. So far he’s played 138 very mediocre games (93 wRC+), after playing 267 phenomenal (albeit lucky) ones to start his career. To me, Gleyber is probably a .260/.350/.410 hitter, which takes out some of the luck he had in his first two years, accounts for the improved plate discipline he has shown over the last two seasons, and posits that there’s no way this power outage is sustainable.

It’s important to note that while Gleyber has regressed in certain areas, like hitting around his xBA, the pendulum has swung very fair in the regressive direction in other areas in ways that appear unsustainable. He’s currently 24 points below his xwOBA, which would tie the greatest variance between wOBA and xwOBA that he’s posted in his career. That’s hard to sustain, and the hot streak that he’s currently on should work toward bringing him back in line. He’s also underpeforming his xSLG by 54 points, which is basically the amount by which he outperformed that stat in 2019. As far as we can tell, neither 2019 Gleyber nor 2021 Gleyber produced (or failed to do so) in sustainable ways. While I don’t see him scaring 40 homers again, it is impossible for me to predict that he will continue to be a black hole offensively. Regression giveth, and regression taketh away.

That final point is really what differentiates Torres and LeMahieu. We have much more reason to believe that Torres is going to rediscover life as a quality offensive player than we do for DJ. DJ has much more data that tell us that he’s basically average. What has average looked like over DJ’s tenure with the Yankees? Eddie Rosario. Manage your expectations. As I said on last week’s podcast, we generally know how good guys are at a sport, and they generally don’t become new players in the middle of their careers. It sucks, but we knew who DJ LeMahieu was when the Yankees signed him. We can’t be all that mad about him playing like that guy.

Candidates for Improvement

Aaron Judge is the best player on the New York Yankees, and teams should be terrified of him finding his mean. Aaron Judge is currently nearly 40 points below his xwOBA, 26 points below his xBA, and more than 70 points below his xSLG. He is posting the second best xwOBA, second best xSLG, and best xBA of his career, despite posting his second lowest wOBA, lowest SLG, and second best batting average of his career. Even with all of this absurdly bad luck, Judge has been 42% better than league average this year.

As always, judge is doing everything right. He’s hitting the ball comically hard, just .4 MPH away from his best average exit velocity. He’s operating close to his career barrel rate, and has tied the best hard hit rate of his career. He is striking out less than he ever has before, though he’s not walking as much as he used to. The only batted ball statistics that really stick out are his slightly depressed fly ball rate and HR/FB rate. A greater proportion of batted balls hit to center field may explain his lower ratio of fly balls to home runs, but really, there’s nothing in his batted profile that suggests Aaron Judge has been the victim of anything but bad luck.

The thing about luck, as we saw from DJ and Gleyber, is that it tends not to last. Bad luck is the same way. Aaron Judge has never finished a season more than 6 points away from his xBA. He’s currently 26 points off. He has never finished a season more than 24 points away from his xSLG. He’s currently 71 points off. He has never finished a season more than 11 points away from his xwOBA. He’s currently 39 points off. Aaron Judge reliably hits a ton of balls in the air very hard. It’s what he does. No matter how well outfielders are positioned or how many balls he hits into the gap that are inexplicably caught, he produces the best possible kind of batted ball (hard hit and in the air) more often than a lot of hitters. He will find his level, and it’s going to be a ton of fun to watch him get there.

The numbers for Gary Sanchez are not nearly as striking as those for Judge, and are likely well within the margin of error for expected stats, but all of his expected stats suggest that he has room to improve without requiring any luck. Gary’s xBA and xwOBA are both about 10 points better than his realized batting average and wOBA, indicating that another hot streak, in which he settles in around .230/.340/.455 as opposed to .216/.329/.450 would not be surprising. His candidacy for improvement is not nearly as strong as Judge’s, but I feel it is my moral duty to point out that Gary Sanchez is still better than you think he is.

Luke Voit is a tough case because of how injuries have derailed his season and given us a very small sample size to work with, but the Luke Voit we are seeing this year is not truly Luke Voit.

Is it possible that this year for Voit is like 2018 for Gary Sanchez? He’s not playing a ton, he’s often hurt when he is playing, and he gets ridiculously unlucky along the way? Sure. But despite what 2018 and 2020 Gary Sanchez tell us, those seasons are hard to have. Like Judge, Voit has played very close to his expected stats over the course of his career. He’s never missed his xwOBA by more than 7 points, and has always been within 25 points of his xSLG, with some more volatility in his xBA. Currently, he’s 125 points short of his xSLG, and 51 points short of his xwOBA. He does not have much time left in the season once he returns from injury to bring his realized stats in line with his expected stats, but given enough plate appearances, it would be hard for the numbers we are seeing to persist.

Parting Shots

The good news on the regression front with the Yankees is that, on the whole, the guys who were bound to regress have done so, and the guys who are bound to improve haven’t yet. Unless incredibly bad luck continues to plague this team, and intensifies, it’s very likely that they already found their floor offensively. They can only go up from here, and it’s reasonable to expect that Aaron Judge and a healthy Luke Voit will be leading the charge when they do. This team can survive DJ LeMahieu and Gleyber Torres not producing the way they did in 2019, and the stats suggest that some of the big boppers are about to go on real tears right when this team needs them.

Enjoy this reel of Judge, Gary, and Voit hitting bombs.

Let’s go Yankees.