Elegy for ’18 – New York Yankees

The Yankees may not have won the World Series in 2018, but they’re set up for a long run of success.
(Photo: Arturo Pardavila III)

Until three of the team’s four postseason starters got knocked out early by the eventual-champion Red Sox, the 2018 Yankees had a successful season. Giancarlo Stanton’s first season in New York may have been a disappointment relative to his MVP 2017, and the rotation required some midseason triage, but the team managed their first hundred-win season since 2009. And thankfully, they did not lose in the wild card game and thus highlight the weirdness in baseball design of combining wild cards with an unbalanced schedule.

The Setup

The early part of the 2017 offseason was wrapped up in the grand hunt for Giancarlo Stanton, a player far more interesting than nearly every free agent actually available for signing. After a number of false starts and mystery teams and trade clauses not-waived, the Yankees came out on top in the race for not-Michael. And unusually when picking up a superstar, it was actually better than simply signing a comparable player in free agency, with the Yankees able to shed Starlin Castro’s salary and only lose two prospects (only one, Jorge Guzman, was a top 10 prospect for the Marlins according to our very own Eric Longenhagen and Kiley McDaniel).

The obsession with the luxury tax ensnared several of the larger-payroll teams over the 2017-2018 winter, and the Yankees were one of the chief actors in this little mini-play. We can argue endlessly about whether the current luxury tax system is well-designed (it isn’t) or whether it serves as a soft salary cap (it does), but it is the system in place and staying under the threshold for a year in order to “reset” the penalty rate provides a tremendous financial motivation to go cheap in the short-term.

The desire to reset the luxury tax penalty heading into an offseason in which Manny Machado, Bryce Harper, and possibly Clayton Kershaw were set to hit free agency was strong, and both of New York’s other significant offseason trades reflected this urge. Chase Headley, a perfectly average third baseman for the Yankees, enough to make him a much better signing than rival Boston’s similar deal with Pablo Sandoval, was sent packing to San Diego in a pure salary dump; the Yankees gave the Padres Bryan Mitchell as compensation for taking Headley’s contract. Any notion that this wasn’t a move designed to trim payroll, that the Padres just really wanted Headley, is undermined by the fact that his new-old team gave him nine starts before sending him to the unemployment line, where Headley spent the rest of 2019.

A three-way trade with the Diamondbacks and Rays netted the Yankees Brandon Drury, who was basically brought in to fill the Headley role of a stopgap until Gleyber Torres or Miguel Andujar; he was made as expendable as a secondary henchman objecting to the antagonist’s devious plot in a James Bond film by the second week of the season.

The Yankees spent all of $14 million on one-year contracts for CC Sabathia and Neil Walker, a far cry from a decade prior, when they guaranteed more than $800 million in contracts after the 2007 and 2008 seasons. Only 2015-2016, when the team’s biggest signing was Chris Denorfia, was quieter.

The Projection

The ZiPS projection system pegged the Yankees as two games better than the Red Sox, with just under a 60% chance of winning the division. ZiPS expected the AL East division title to essentially be yet another Yankees-Red Sox battle, with only a 4% chance of one of The Others of surprising enough to take the division. ZiPS was confident about the Yankees’ offense, seeing most of the unknown as a matter how quickly Andujar and Torres would have full-time jobs and how effectively the Yankees would continue to yank Jacoby Ellsbury’s playing time. The bullpen was projected to be the best bullpen that ZiPS ever projected. The computer’s main worry was the back of the rotation, which the computer did not see as very deep should something happened to one of Sabathia, Sonny Gray, or Jordan Montgomery.

The Results

Oddly enough for a team that won 100 games, it felt like the Yankees had more than their fair share of disappointments. Some of the fears about the rotation came to pass; Gray’s command was a tire fire in the first half and Montgomery’s season — and most, if not all, of 2019 — ended in June with Tommy John surgery. Neither Domingo German or Luis Cessa proved to be ready for a rotation spot on a win-now team, and the surprising Jonathan Loaisiga was yanked from the rotation with shoulder pain, leaving the team with obvious back-rotation holes going into the trade deadline.

Unlike a team like the Giants, who could never have made a significant midseason addition without going over the luxury tax threshold (they only had a $300,000 cushion at one point), the Yankees left themselves some space to make move that would require them taking on salary. It was enough space that the team was able to add Lance Lynn and J.A. Happ for the stretch run, and pick up Zach Britton from the Orioles to make a deep bullpen even deeper. None of these moves ended up changing the team’s postseason fate, as New York fell short in the contest for the division, but they might have if the team had gone deeper into the playoffs than they actually did.

As projected by ZiPS, the team set a new all-time record for team home runs in a season with 267, though to be honest, that result wasn’t particularly surprising. But even the second-ranked scoring offense in the AL has some plans go awry. Gary Sanchez, who had established himself as a star-level catcher in his first 1 1/2 seasons in the majors, lost a hundred points of batting average, finishing at .186/.291/.406 (he was Rob Deer-like in that he still was worth 1.4 WAR in 89 games). Further marring his season was the charge that he lacked hustle, which, combined with a groin injury, led to weeks of conspiracy theory about his health status.

Stanton also has to be considered at least a mild disappointment, dropping to 38 home runs and a 127 wRC+ from 59 and 159 his final year with the Marlins. Now, it would be greedy to focus too much on this dip — complaining about a 4.2 WAR player is a high class problem to have — but the fact remains that the Yankees did not get as much from their newly acquired star as they would have liked to see. Greg Bird managed to stay healthy for the second-half of the year, but also managed to stay around replacement level, resulting in him mostly losing his job to Luke Voit.

Those disappointments, even when combined with the Brett Gardner starting to show his age, turned out not to really matter. Aaron Hicks can rightly be described as a legitimate All-Star, which still seems a little strange to 2016 Dan, but that’s the world we’re in now. Andujar and Torres finished second and third in the Rookie of the Year voting (I would have flipped them given Andujar’s poor defense). Aaron Judge’s regression toward the mean indicated his mean was pretty damn high.

One interesting note is that ZiPS never actually knocked the Yankees down behind the Red Sox in projections. Even with the eight-game cushion at the end of the season, ZiPS still saw the Yankees as a sliver better than the Red Sox, though you wouldn’t have known it from their four-game playoff series.

What Comes Next?

In the early offseason, the Yankees have played the “Golly gee, I don’t know, the root cellar needs a’fixin’ and I’m not sure we have the money for those big city fancies with grandpa’s water on the knee” card publicly when it comes to the cream of this year’s free agent crop. This is hardly unusual this winter; most of the big spending teams, including the Red Sox, Dodgers, and Cubs, have all been mumbling this storyline with only a few variations on the theme. Only the Phillies, with their talk of “stupid money” have really broken ranks.

That’s not to say the team has done nothing, but the moves they’ve made have largely been keeping the band together. Gardner and Sabathia, two primary remaining holdovers from the team’s prior core, will return in 2019 on one-year deals. Happ, who stabilized the back of the rotation in late summer, returns in that role for two more years.

The team also made one of the bigger trades this winter, picking up James Paxton from the rebuilding Seattle Mariners for a package led by Justus Sheffield. With a rotation that now looks like Luis SeverinoMasahiro Tanaka-Paxton-Happ-Sabathia in 2019, I think at least when it comes to the pitching, the Yankees will have a quiet rest of their offseason.

Otherwise, I’m not so sure that the impression the team has given of only dipping their toes into free agency is just posturing. Ten years ago, I’d have cried total bull, but with even large-market teams seeming generally less interested in splashing cash than they have been at any point I can remember as a baseball fan (I’m 40), I’m not really sure right now. Bryce Harper or Manny Machado ought to be a fit, as would someone like Brian Dozier to fill-in at second with Torres presumably at short while Didi Gregorius recovers from surgery, but I just don’t know if the team’s hinted lack of interest is genuine or not.

There’s a bit of a prisoner’s dilemma going on with the Red Sox and Yankees, both teams that ought to be in the top three in MLB in 2019, in that both of them spending $200 million might not advantage either over a scenario in which both spend very little. What actually happens is one of the most interesting questions remaining this winter. The Yankees will be a very good team in 2019, but I’m quite unsure how much they’ll open up their ceiling this offseason.

ZiPS Projection – Giancarlo Stanton

How much will Stanton bounce-back from a weak-ish 2019? How beneficial would it be for him to opt out after 2020? How high can he get in the all-time home run rankings? These are questions, naturally, for the ZiPS-o-matic 5000.

No, I’m not actually calling it that.

ZiPS Projections – Giancarlo Stanton
Year BA OBP SLG AB R H 2B 3B HR RBI BB SO SB OPS+ DR WAR
2019 .255 .344 .557 564 98 144 27 1 47 121 72 198 4 135 5 4.6
2020 .254 .344 .563 544 96 138 28 1 46 119 72 195 3 137 5 4.5
2021 .251 .343 .550 533 92 134 28 1 43 113 71 187 3 133 5 4.1
2022 .255 .346 .555 517 90 132 27 1 42 111 68 174 3 135 4 4.1
2023 .249 .341 .538 498 84 124 25 1 39 102 66 165 3 130 4 3.6
2024 .245 .335 .510 478 76 117 23 1 34 92 61 152 3 122 3 2.9
2025 .240 .326 .482 454 69 109 21 1 29 81 55 137 3 112 3 2.1
2026 .235 .317 .452 429 60 101 19 1 24 71 48 120 2 103 2 1.3
2027 .230 .306 .421 378 49 87 16 1 18 57 38 96 2 92 2 0.6
2028 .222 .290 .381 257 30 57 9 1 10 34 23 59 1 78 1 -0.1

ZiPS is more negative on Stanton than I had expected. It isn’t thrilled by his step backwards in plate discipline from 2017, now seeing Stanton with a higher chance at going down that “old player skills” career path than establishing a high-enough level for a more graceful decline phase. A lot of players who didn’t age particularly well have crept up in his similarity group, with names like Rudy York, Jack Clark, Jay Buhner, Richie Sexson, and Boog Powell all in the top ten. That gets Stanton up to 637 home runs, but like Pujols, has him petering out before he seriously gets into the Ruth/Aaron/Bonds battle.





Dan Szymborski is a senior writer for FanGraphs and the developer of the ZiPS projection system. He was a writer for ESPN.com from 2010-2018, a regular guest on a number of radio shows and podcasts, and a voting BBWAA member. He also maintains a terrible Twitter account at @DSzymborski.

35 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Roland
5 years ago

Judge missed seven weeks with a broken wrist. That’s regression towards the mean?

Roland
5 years ago
Reply to  Dan Szymborski

Production isn’t linear Dan. You would think you would lead with the injury, its impact for the remainder of the season and maybe that he was one of the most negatively impacted players in mlb in balls called strikes against him.

Eminor3rdmember
5 years ago
Reply to  Roland

You seem to be searching for a narrative that simply was not included or intended in the article.

Roland
5 years ago
Reply to  Dan Szymborski

Dan, your second paragraph isn’t clear to me.

I don’t see how you can discount the fact that Judge missed many of the worst teams in the AL during August.

Also, Judge on 9/18/17 – 275/413/584
9/18/18 – 282/395/543

Yes it’s “regression”, it’s also merits some discussion. The livelier ball in 2017 is something to consider.

Roland
5 years ago
Reply to  Dan Szymborski

I think your premise is flawed. Look at the pitchers OPS for the stretch Judge missed, then get back to me.

sadtrombonemember
5 years ago
Reply to  Roland

His ISO went from .343 to .249, and his overall offensive value (as measured by wRC+) fell as a result.

Some of that was the 51 PAs after he returned from his injury, where he didn’t have his power back, because wrist injuries are hell on hitters. Most of that was that his power output declined across the board; he didn’t have a single month with an ISO as high as his overall ISO the previous year.

Roland
5 years ago
Reply to  sadtrombone

Judge missed the entire month of August and the softest part of the Yanks schedule as well.

It’s an incomplete season. To merely state it was “regression” is lazy and inaccurate.

sadtrombonemember
5 years ago
Reply to  Roland

But…it WAS regression. Just because he missed time doesn’t invalidate the point that his rate stats regressed.

HappyFunBallmember
5 years ago
Reply to  Roland

Injury can cause regression. So can too many hangovers, divorce, bad luck, lack of previous good luck, or lazy workout habits. Regression is a result not a judgement.

emh1969
5 years ago
Reply to  HappyFunBall

Actually it’s not regression to the mean because we don’t have enough yearly data points to know what Judge’s mean is. Throwing out his lousy first season, all we have are two data points, a wRC+ of 172 and 149. Now it could be that Judge’s potential performance ranges from 140-200 wRC+ with a mean of 170. In which case his 2017 season represented an average performance and last yer represented a below average performance.

Dan seems to be talking about regression to the league mean but I’m not sure why you’d talk about it that way. Players regress to their own mean, not the league mean.

sadtrombonemember
5 years ago
Reply to  emh1969

This is technically true, but in this case it is still unlikely that his true talent is 170 wRC+. And that is because of the distribution of wRC+, and how many players wind up at that extreme.

My preferred way of thinking about your comment (heuristically) in relation to our “unknown unknowns” would be that it is possible that the true talent could lie somewhere in between 149 and 172. Thus, his regression could have been a mixture of regression to the mean and other factors.

That said, technically speaking we don’t know Judge’s true wRC+ and your example is technically as valid as mine.

The more accurate way of rephrasing is probably: “Judge’s regression in 2018 is very likely at least partly a regression to his own personal mean since the odds that anyone has a true-talent wRC+ of 170 is exceptionally low, but we need more data to be sure.”

Johnston
5 years ago
Reply to  sadtrombone

“Judge’s regression in 2018 is very likely at least partly a regression to his own personal mean since the odds that anyone has a true-talent wRC+ of 170 is exceptionally low, but we need more data to be sure.”

Precisely and very well put.

emh1969
5 years ago
Reply to  Dan Szymborski

Honestly, though we shouldn’t even be talking about regression from one year to the next. Any player’s “true talent level” at age 26 is different than what it was at age 25. Not to mention that there are extraneous factors from year to year that can affect performance, regardless of talent level.

emh1969
5 years ago
Reply to  emh1969

And yeah, I agree that it’s unlikely that Judge’s true talent at either age 25 or age 26 is a 170 wRC+. But we have no way of knowing that for sure. Last year, when Judge got off to a hot start and had a 178 wRC+ on May 17th, Travis wrote an article saying that Judge was hitting better despite having a worse strike zone. At that point, people seemed to just take it for granted that Judge’s true talent level WAS around a 170 wRC+ (I was one of the few naysayers on the article, pointing out that Judge’s “improvement” was a statistical artifact of the Yankees having played 25 of their 41 games at home. I was voted down in spades for that comment but ultimately was proven correct).

sadtrombonemember
5 years ago
Reply to  emh1969

One of my favorite things that Jay does (among the many cool things he writes), is that when he writes stuff up like that, is he calls it “small-sample the theater.” I think that helps bracket the discussion a bit, that this might be a breakout, or it might just be a guy who got hot for a little bit. Assuming a stable underlying true performance level, you would expect more extreme performance (both good and bad) in smaller samples because there aren’t enough “trials”.

I went back and looked up that article because I was curious. Holy hell! I missed that one, and I’m glad everyone is picking up the slack critiquing the hell out of “clutch” and “WPA”, which range from bar-trivia-relevant to actively-misleading depending on the context.

emh1969
5 years ago
Reply to  sadtrombone

“is he calls it “small-sample the theater.””

Um yeah, that’s MY point. If we’re talking about season-based performance, then we have only two data points and we shouldn’t be making any conclusions based on an n of 2.

Of course, Judge is such an extreme player that we should take that into account. In 2017, he put up a wRC+ of 198 in 334 PAs at Yankee Stadium. Now Judge couldn’t possibly have a true talent level of nearly 200 wRC+ at home, could he? Well in 2018, he actually bested that, putting up a 210 wRC+ in 244 PAs. So yeah, it seems like his true talent level at Yankee Stadium is somewhere in the range of 200 wRC+.

So one thing we should acknowledge is this: while in general we shouldn’t expect a player to have a true talent of 170 wRC+, if said player has a true talent level of 200 wRC+ at home, then a 170 wRC+ overall true talent level all of a sudden becomes a LOT more possible.

Which means, to the extent that he “regressed”, it happened in his road games. On the road, he went from a 147 wRC+ to a 98 wRC+. That’s a pretty substantive decline. And the numbers are different enough from one another that it’s hard to draw any conclusions abut his true talent level on the road. (Not to mention that we’re still dealing with an n of 2). Is it 147? 98? Somewhere in-between? There’s no way to really know but I’ll bet it’s a LOT closer to 147 than it is to 98.

Roland
5 years ago
Reply to  Dan Szymborski

Since Mike Trout has a career 172, I’d say that’s a safe assumption. That is an incredibly high bar to meet.

I still believe his regression is due to the missed third of the season, against the worst competition and during the period when the ball carries the most, plus the 8.5% decline in HR from 2017-2018.

If they keep the ball less lively, no one may see ISO above .340 again.

That said, his 2019 projections seem overly negative.

Johnston
5 years ago
Reply to  Roland

“I still believe his regression is due to the missed third of the season“

Roland, you may believe that, but every fact about his season says otherwise. I can well understand being a partisan fan of a player – I was such a fan of Ozzie Smith myself – but you need to look at the data, and the data clearly says that what you believe just isn’t so.

Sorry, but there it is.

rhswanzey
5 years ago
Reply to  Roland

I think it’s worth mentioning that projection systems are a model. If you went through each player and accounted for injury, personal problems, micro-level schedule and quality of competition: (1) you’d round up a lot of players at the expense of the overall accuracy of the model; (2) we wouldn’t get ZiPS projections because it would take thousands of labor hours to make all those subjective adjustments.

If you’re about to sign or trade for Judge, of *course* you’re pouring into all the data available and looking hard at injury history. To my knowledge, that’s not what a leaguewide projection model is designed to do. If you went through and did this with any given name player, you’re probably not going to beat the model’s accuracy, and if you can, you should design your own.