<
>

Baseball gods or merely human? Four stars projections can't predict

How well will Giancarlo Stanton fare in the Bronx? How about a full year from Justin Verlander in Houston? The three major projections for 2018 are out -- and, in a rarity, they totally disagree on four star players. So who settles the argument? You do! Kim Klement/USA TODAY Sports

One hypothesis for baseball's sluggish free-agent market goes that, in the post-Moneyball era, front offices have "narrowing differences in player valuation." In addition to their scouts, many -- if not all -- teams have their own projection systems, and projection systems tend to follow similar logic and view players in similar ways. For example, here's what to expect from Red Sox sophomore Andrew Benintendi this year, according to three public projections:

  • ZiPS: 18 HR, .282/.354/.456, 84 R, 17 SB, 2.9 WAR

  • Steamer: 19 HR, .286/.360/.464, 88 R, 15 SB, 2.8 WAR

  • PECOTA: 18 HR, .272/.345/.446, 86 R, 16 SB, 2.5 WARP

Three systems and three methodologies, but ask them to predict 600 plate appearances, and they disagree on only a half-dozen of them. Most players' projections aren't quite this uniform, but rarely will you find yourself flabbergasted by a projection. Nor, probably, should you be, if you've been paying attention to the player's performance.

But this makes those few instances of system disagreement especially interesting. There are, occasionally, players whose outlooks look significantly different depending on the URL at the top of your browser. In turn, those projections might dramatically affect an entire team's outlook or persuade you to push a player six rounds higher in a fantasy draft. These players are windows into the science of projections, and they're helpful reminders of the utter bonkersness of baseball performance. Settle a few of these disagreements for us, will you?

Giancarlo Stanton

  • ZiPS: 6.2 WAR, second-best position player in baseball

  • Steamer: 5.8 WAR, fifth-best position player in baseball

  • PECOTA: 4.5 WAR, 14th-best position player in baseball

The difference between second- and 14th-best might not seem like much, but it's the difference between Aaron Judge and Anthony Rendon last year, a difference I certainly noticed.

ZiPS' and Steamer's disagreement is over defense and role: Stanton is worth less as a DH, where he'll likely play some games. Both see Stanton as the second-best hitter in baseball, behind only Mike Trout. ZiPS sees him hitting 55 home runs with a 1.033 OPS; Steamer sees 53 homers and a 1.022 OPS. That's the same basic hitter with small disagreements over how much he contributes with his glove.

But PECOTA is much more conservative on Stanton the hitter, projecting 41 homers (still the most of any projected hitter) and a .908 OPS. For this, it's important to remember that in 2016, Stanton hit just .240/.326/.489, and to that point in his career, he had never hit 40 home runs, he had a sub-.900 career OPS, and he had dealt with significant injuries in four of the previous five seasons.

It's also important to remember that baseball is a cruel game that occasionally flummoxes even the very best players. Aware of that potential for disaster, PECOTA often looks conservative: It's folding in the very real possibility of collapse for each player. In fact, it often looks too conservative -- every year, Mike Trout is projected to have a career-worst season, and every year he bests it -- but collectively, the caution seems to pay off. Over the past three seasons, PECOTA's top 20 projected hitters -- 60 hitter seasons in all -- have collectively done exactly what PECOTA projected for them, even though more have outhit their projections than come up short.

That gives us three visions of Stanton: The AL's MVP, a candidate for MVP but a little bit down from 2017, and a valuable All-Star who nevertheless comes up well short of preseason hyperbole. Interestingly, the Fan Projections at FanGraphs -- which are based on readers' crowd-sourced estimates -- lean toward the pessimistic view. That's very rare; fans are almost always more optimistic than soulless projection systems.

Jose Bautista

For six years in a row, Bautista was an All-Star. Does that seem that long ago to you? Do his top-10 MVP finishes, does his bat flip in the ALCS, seem forever ago? If they do, you're ZiPS. If they don't, you're PECOTA.

  • ZiPS: -0.1 WAR, .699 OPS, 582nd-best position player in baseball

  • PECOTA: 3.0 WAR, .812 OPS, 55th-best position player in baseball

  • (Steamer doesn't have a WAR because it doesn't estimate playing time for players who are still free agents. It splits the difference on his offense, though: a .761 OPS.)

Remember when we said projection systems tend to follow similar logic? The basis for just about every projection system is what the player has done, how recently and at what age. Some systems go further back for information; some put greater or lesser emphasis on recency; some have different aging curves; some have different ways of neutralizing performances based on ballparks or quality of competition; some fold in other details, the sorts of details more traditionally associated with scouting; some draw more inference based on previous, similar players. But the basis is simple: What has he done, how recently and at what age?

Over the past three years, Bautista has been a very good hitter: He has hit more home runs than George Springer, has a higher OPS than Wil Myers, has outhit (by weighted on-base average) Marcell Ozuna and Justin Smoak. That's all really good, and it's relevant. Good players have bad years, often followed by more good years. But last year, Bautista was one of the worst hitters in the game, and he's old, and those six consecutive All-Star appearances might not tell us much about what's going on in his cells right now. As a result, one system sees a player who was extremely good not long ago. Another sees a player who was extremely bad very recently.

Trea Turner

Turner, the Nationals' 24-year-old shortstop, has never made an All-Star team, never received an MVP vote, never led his league in any stat, never even qualified for the batting title. (He was called up late in 2015 and 2016, and he broke his wrist in the summer of 2017.) But he's already a star -- or at least close to it. He has been a top-five pick in fantasy drafts this spring, with 20-HR power and 50-SB speed. ZiPS projects him to the 37th-best position player in baseball, with his top player comp being Hall of Famer Ryne Sandberg. Steamer projects him to be the 22nd-most valuable, just behind Freddie Freeman and just ahead of Judge. Impressive.

But then PECOTA butts in. It projects Turner to be the third-most valuable player in baseball, at 6.1 WAR. Surprisingly, it doesn't forecast much better offensive numbers:

  • ZiPS: 3.0 WAR, .774 OPS

  • Steamer: 4.1 WAR, .813 OPS

  • PECOTA: 6.1 WAR, .805 OPS

The bulk of the gap comes from smaller or harder-to-measure or easier-to-disagree-on things. PECOTA predicts that Turner will be, as a baserunner, eight runs better than the average runner. Steamer and ZiPS say only 3.5 runs and 5.3 runs, respectively. That accounts for about a quarter of the difference between Steamer and PECOTA.

PECOTA also projects Turner to be one of the very best defensive shortstops in baseball, at 11 runs above average. ZiPS and Steamer project him to be just one or two runs above average. Each projection uses different advanced metrics to get to a number, but Turner is especially hard to project as a shortstop because he has played second base, center field and only 103 games at shortstop in his brief major league career. PECOTA is likely drawing from his minor league performance at the position, which (it thinks) was quite impressive.

Honestly, the public might hardly notice the difference between these three projected outcomes. In all three cases, he'd help your fantasy team the same. But the Nationals would notice. Turner's baserunning and defensive value could span the distance from top MVP candidate to being left off most ballots.

Justin Verlander

According to Baseball Prospectus' advanced pitching metric, deserved run average -- which considers a dizzying array of factors in and out of a pitcher's control -- Verlander was the eighth-best pitcher in baseball last year. DRA didn't just like him after the trade to Houston, when his ERA was 1.06 in five starts; it thinks he pitched just as well in Detroit, despite a more mediocre ERA. In 2016, DRA says, Verlander was the best pitcher in baseball.

Which makes this one surprising:

  • ZiPS: 4.0 WAR (16th among all pitchers), 3.43 ERA, 1.12 WHIP

  • Steamer: 3.8 WAR (19th among all pitchers), 3.86 ERA, 1.20 WHIP

  • PECOTA: 2.2 WAR (48th among all pitchers), 4.18 ERA, 1.29 WHIP

This probably comes down to how each system treats aging. Darius Austin, at Baseball Prospectus, wrote this month about the pessimistic case for Verlander:

"While we might look at Verlander and see a workhorse who has a significant track record of success -- including back-to-back seasons over 200 innings with more than a strikeout per inning and a very recent spectacular playoff performance -- PECOTA sees a 35-year-old with a walk rate over 3.0/9, a home run problem, and a fairly recent season in which he was no better than league average. Most sobering are Verlander's top three same-age comps: Adam Wainwright, Jason Schmidt and the late Roy Halladay, all reminders of how rapid the decline can be."

Of course, nothing about Verlander looks to be in decline. After decreased velocity led to poor 2014 and 2015 seasons -- which pollute his projections somewhat -- he regained his physical dominance. His average fastball last year was harder than it had been since 2010 and 2.5 mph harder than in 2014. Plus, another of his top same-age comps is Roger Clemens.

The punch line is that Andrew Benintendi will probably do something entirely different from his projections this year. Projection systems might agree, or slightly disagree, on what a player's most likely outcome is. But one assumption baked into all of them is that the range of outcomes is fantastic. Some part of ZiPS knows PECOTA might be right about Bautista; some part of PECOTA knows Steamer might be right about Stanton; and some part of Steamer knows the only thing we can really count on is looking back at what we thought we knew and laughing.