One of my favorite things about Statcast is how it quantifies hitters’ contact in such a way as to indicate what the results should look like, most succinctly through xwOBA, the contact-quality-based equivalent of the all-encompassing hitting stat, wOBA. This is helpful for discerning when a guy’s had a lot of screaming line drives caught, or when a guy’s had a lot of routine groundballs find a hole. You’d think it would be rather predictive of a player’s performance going forward. And it’s not not predictive of that. But the data’s fuzzy.
Today, we went through the 2015-2019 seasons (2015 was the beginning of the Statcast era, we decided to exclude 2020 because of the pandemic-shortened schedule) and compared the wOBA’s and xwOBA’s across seasons for qualified hitters, specifically measuring how predictive a hitter’s wOBA and xwOBA were of their future results. Here’s what we found:
| Predictor/Predicted | R-Squared |
| wOBA/wOBA | 0.31 |
| xwOBA/wOBA | 0.26 |
| xwOBA/xwOBA | 0.51 |
These aren’t terribly low r-squared values. They’re much higher than what we got when comparing ERA, xERA, and FIP back this spring. But it’s noteworthy that a hitter’s actual results are more predictive of his results the next year (should he again reach the qualifying number of plate appearances) than his expected results are of those same actual results.
That said, xwOBA is much more predictive of future xwOBA than wOBA is of future wOBA. Contact quality stays consistent over time more so than results do.
The takeaway? I don’t have a big one. Two possible explanations, though, are that 1) defensive positioning—speaking more broadly than “the shift”—really affects the relationship between contact quality and results and 2) xwOBA might not be that good, or 2b) xwOBA might not be that good yet, with only a few years of data in the system so far. Would require a deeper dive to check.
Anyway, good to know when evaluating hitters. Interesting too how much more consistent hitters’ performance is than pitchers’ year-over-year.
