The fantasy community is down on Justin Upton. I get it, but it’s a little strange to me given our collective penchant for recency bias. Upton had a monster second half and finished the season an almost-perfect replica of his usual self. (The operative qualifier being “almost.” We’ll get to that in a second.) Sure, it was a rocky year, but hey, Joey Votto had one, too. Dude was batting .213 with a 27 percent strikeout rate (K%) through May…

Right, so Upton was an almost-perfect replica of himself. In a vacuum, his production looks nearly identical to his typical annual accomplishment, down to nearly every statistic except for his batting average on balls in play (BABIP). In my investigation of his woes, I noticed his uncharacteristically low infield hit rate (IFH%). Here’s a list of hitters with higher infield hit rates than Justin Upton in 2016:

• Salvador Perez
• Kendrys Morales
• Albert Pujols
• One hundred and thirty five other players not surnamed Ortiz, Mauer, Seager, Granderson, Cabrera, Votto or Kemp

Yes, Upton ranked among the bottom 6 percent of hitters in terms of infield hits. If there’s a single bone to pick about Upton’s season — well, aside from the insane volatility — it’s that his BABIP failed to get back on track, continuing to linger at a league-average mark. It seems a trend has emerged; accordingly, it’s easy to accept said trend as a new normal, as a resignation of Upton’s fifth tool.

I’m here to make the classic* Infield Hit Rate Defense, or IHRD, as it’s known in the infield hit community.**

*Not classic
**Not a community

Infield hits likely have an inherent skill to them. The average ground ball has a 24 percent chance of becoming a hit. How many ground balls to the outfield go for hits? Probably all of them, except for, like, one or two, tops. And, inversely, how many infield ground balls go for hits? Probably very few. (Six percent of all ground balls, to be exact, or roughly 7.4 percent of infield ground balls.)

An infield ground ball by its very nature can be fielded cleanly, so a successful infield hit would require a non-trivial amount of speed to leg it out. Per Bill James‘ speed score (Spd), Upton was barely below average (4.2 to the league’s 4.4) in 2016. It’s not unreasonable to think he should have recorded a roughly league-average number of infield hits. Not coincidentally, his 6.9 percent infield hit rate during the previous half-decade seem to validate such a claim.

That’s not a small number of hits, mind you. Upton hit 152 ground balls last year. The difference between rates of 2.5 and 6 percent is roughly five hits, or eight points of batting average; 7 percent, and it’s 11 points of average. It’s not much — we are absolutely splitting hairs here — but it would have bumped him up about 10 spots in ESPN’s Player Rater.

Frankly, I simply wanted to use Upton as the premise for investigating infield hits, something I have never really seen discussed (mostly because it’s such a benign topic). Namely, I was curious as to how infield hit rate (and changes to a hitter’s infield hit rate) track from year to year using a variety of linear regression specifications:

(1) Current-year IFH% regressed against previous-year IFH%

IFH%_t = 0.5601*IFH%_t-1 + 0.0269
Adjusted R²: 0.33

The simplest approach: the prior year’s infield hit rate regressed against the current year’s rate. Implication: Any hitter with legs should record infield hits on ground balls about 2.7 percent of the time — more often than Upton, who decidedly has not one but two legs, did last year.

Expected IFH: 5.9%

(2) Current-year IFH% against previous-year IFH% and age

IFH%_t = 0.5427*IFH%_t-1 – 0.0045*age + 0.000064*age² + 0.1038
Adjusted R²: 0.34

Hitters likely grow slower as they get older, so I incorporated a nonlinear age element. It produces a slightly better fit than the first.

Expected IFH: 5.4%

(3) Change in IFH% from Y₁ to Y₂ against change in IFH% from Y₀ to Y₁

ΔIFH%_t = -0.4840*ΔIFH%_t-1 – 0.0032
Adjusted R²: 0.24

To economists or statisticians, this might be a more theoretically sound way of developing a predictive model. (It helps dispel concerns of autocorrelation.) It’s less intuitive, perhaps, but what it does is finds the correlation not of the infield hit rate itself but, rather, the margin by which it changes from one year to the next. So, instead of regressing Upton’s 2.5 percent in 2016 versus his 10 percent in 2015, it regresses his 7.5 percent decrease from 2015 to 2016 against his 4.0 percent increase from 2014 to 2015. For all players across all years! These kinds of models naturally produce lower correlation measures, but it does not make them inferior. No matter what you think of the approach, it really helps characterize regression to the mean, that’s for sure: The coefficient (-0.4840) on ΔIFH%_t-1 indicates that any change from a previous year should bounce back almost exactly halfway (48 percent of the way, to be specific).

Expected IFH: 5.8%

(4) Current-year IFH% against previous three years’ IFH%

IFH%_t = 0.7281*IFH%_t-1→t-3 + 0.0137
Adjusted R²: 0.44

Note the R². It’s well-established that a player’s previous three years of performance can predict his forthcoming season more adequately than his most recent season alone. It shouldn’t overrule news and validation of swing adjustments, changes in repertoire, and all that, but it should help you overcome your recency bias a bit for dudes with fluctuations to their established track records.

Expected IFH: 6.4%

Average expected IFH: 5.9%

…which is pretty much exactly what I intuited.

So, maybe Upton’s BABIP, whether deserved or undeserved (because who’s to say an infield hit is truly deserved but also not random?), should be a dozen-or-so points higher. That gives me a little more hope. Applying method (4) to his pop-up rate (IFFB%*FB%) yields another two hits for roughly 15 points of missing batting average in all.

This wasn’t meant to be anything profound — mostly just an excuse to satiate my curiosity. Players suffer bad luck on all types of balls in play, not just on infield ground balls, every year. Nor do I expect to assuage the fears of those whom Upton tormented. But at 81st overall per NFBC ADP, and being one of only 11 players to finish in the top 100 the last five years, I’m bullish on J-Up, for not only his a rebound but also his proven reliability — something that’s not as easy to come by as one might think.