Hard%, xwOBA, and the De-Juiced Ball

The league-wide hard-hit rate (Hard%) is up. Like, way up, at its highest level by far in the 17 years Baseball Info Solutions has measured and tracked the statistic.

Yet league-wide home runs are down, and way down, too, not in the whole history of the game but at least in the context of the recent Juiced Ball Era^TM. Hard-hit rates and power, as measured by home runs or isolated power (ISO), increased steadily and in tandem from 2015 through 2017. You’d expect, then, that if the ball were still juiced in 2018, the league’s highest hard-hit rate ever might produce the highest league ISO ever.

No such luck, though; 2018’s .161 ISO falls a full 10 points short of last year and a tick short of 2016. Which is odd, see, because batters are hitting the ball harder than ever. Since 2015, when sabermetricians first noticed the ball was juiced…

• Average exit velocity (EV) is at its highest
• Average EV on fly balls and line drives is at its highest
• Barrels per batted ball is at its highest
• Ideal hits* per batted ball is at its highest
  *EV > 96 mph, launch angle between 21 and 36 degrees

… and all by substantial margins. Yet, at the same time…

• Home runs per barrel is at its lowest
• Home runs per barrel hit 100+ mph is at its lowest
• Home runs per ideal hit is at its lowest
• Average distance of an ideal hit is at its lowest

… and all, again, by substantial margins. The inputs no longer match the outputs, something Rob Arthur, baseball’s preeminent juiced ball researcher, noticed as early as the second week of April. It’s something the community at large tried to attribute to unusually cold weather.

It’s substantially clear now — to me, at least — the ball has been de-juiced. (I don’t really care for semantics or the official prognosis, from MLB or otherwise; it waddles like a duck and it quacks like a duck.) I’m surprised we all didn’t take this Sports Illustrated report from February about league-mandated humidors more seriously, or maybe I blacked out for several weeks while on Twitter this preseason (I didn’t even know it existed until June!). Anyway, the best we can do is digest what information we have to make it usable.

I regressed hard-hit rate against isolated power using year fixed effects for all qualified hitters since 2002 (n = 2,566), which produced an adjusted r² of 0.64, indicating the two metrics are very highly correlated. The model resoundingly indicate a hard-hit rate of, say, 30% in 2018 isn’t the same as a hard-hit rate of 30% in 2017. In fact, it’s much lesser than: holding contact quality constant, ISO is down nearly 40 points from 2017 and 30 points from 2016. Or, stated inversely, hitters need to make hard contact 4.7 percentage points more frequently in 2018 to produce a 2017-equivalent ISO.

The same applies to home runs per fly ball (HR/FB), too (adjusted r² = 0.60): HR/FB is down roughly 4 percentage points, all else equal, and hitters have to make hard contact 5.0 percentage points more frequently to replicate a 2017-equivalent HR/FB rate.

That’s kind of a big deal. It doesn’t mean you should dismiss hard-hit rate as a usable metric; it still very much is. And you should still celebrate hitters who improve their hard-hit rate year over year. It’s just that gains are substantially less meaningful in 2018 — and losses are all the more detrimental. What’s important to note is Hard% isn’t broken. It has peaked in 2018 just as exit velocity and barrels and ideal hits have all peaked. From where I’m sitting, it is still characterizing contact quality commensurate with Statcast’s raw and stylized measurements of contact quality.

In addition to the aforementioned ISO and HR/FB implications, there seems to me a profound Statcast-related implication as well. Statcast calculates expected weighted on-base average (xwOBA), which has become a go-to metric for the fantasy baseball community for anticipating regression. In theory, any “expected” metric predicated upon outcomes, in aggregate, should equal the actual metric it is estimating, such that league-wide wOBA equals xwOBA (just as league-wide FIP and xFIP equal ERA). It is not surprising, then, that the league-wide differential between wOBA and xwOBA (wOBA-xwOBA) on barrels is +0.002 — almost, but not quite, perfectly zero.

What is surprising, however, is xwOBA differentials for barrels vary wildly from year to year:

• 2015: +0.069
• 2016: +0.033
• 2017: +0.010
• 2018: -0.157

xwOBA is dramatically overestimating wOBA on barrels in 2018, and you can bet your buns it relates to the ball being de-juiced. It reveals the problem inherent to xwOBA: it doesn’t control for what might be immense exogenous differences in any given season such as, for example, the ball being juiced and then de-juiced. Two balls hit at X mph, Y vertical angle, and Z lateral angle — one in 2017, one in 2018 — will theoretically have the same xwOBAs. That’s a problematic assumption if one ball is juiced in one season and de-juiced in another.

Hard% doesn’t correlate as strongly with xwOBA on barrels and solid contact (adjusted r² = 0.33), both of which are Statcast’s most comparable contact quality classifications of the six it provides. Yet the trend persists: hard contact in 2018 is worth less than hard contact in 2017, all else equal. The lack of year-specific controls effectively makes xwOBA blind to the physical differences of the baseball.

Granted, hard contact comprises only a third of all batted balls and, consequently, even less of all plate appearances. Thus, the deficit related to barrels and solid contact has a relatively small impact on overall xwOBA differentials, but it’s an impact nonetheless:

• 2015: +0.005
• 2016: +0.002
• 2017: 0.000
• 2018: -0.017

The differential in 2018 will likely shrink as the season trudges onward, but it’s unlikely it will flip by October unless MLB discontinues its mandated use of humidors (or maybe it has always been about the seams).

Hitters are making hard contact more frequently yet hitting for less power. Hard-hit rate isn’t broken, but xwOBA might be: it (seemingly) does not yet control for the de-juiced ball, making your favorite hitter liable to underperform his xwOBA in 2018. Take care to calibrate your analytical toolset accordingly.