The Z Files: Is Average Exit Velocity on Groundballs Useful?

Increased use of advanced metrics for fantasy baseball analysis is obviously great. However, there's a downside. As more incorporate the data into their work, pressure mounts for others to follow suit. Unfortunately, the repercussion is not everyone fully understands the derivation and application of all that's out there, resulting in misuse.

A prime example of this is the expected Statcast stats, such as xBA (batting average), wSLG (slugging percentage) and xwOBA (weighted on base average). It's become commonplace to compare actual results to expected, looking for significant deltas. An actual stat higher than expected is often deemed lucky, and vice versa. On the surface, this makes sense. However, there's a major flaw in that reasoning.

Statcast deploys a unique means of measuring hit probability. In short, the components (exit velocity, launch angle, sprint speed, etc.) are compared to a database of similar batted ball outcomes and assigned a hit probability. What percentage of similarly batted balls are hits? What is the distribution in terms of single, double, triple and homer?

By means of example, if a specific batted ball was a hit 50 percent or the time, it counts as 0.5 hits and gets included in xBA as such. If it cleared the fence 10 percent of the time, it's logged at 0.1 HR and gets factored into xSLG, xwOBA etc, in that manner.

Missing from the above description is park influences. The same traits could be a homer 80 percent of the time in Yankee Stadium only to be caught at a 70 percent clip in Oakland. Especially on lofted balls in play, expected stats can be misleading. I'll be talking more about this down the line.

Another potentially misleading element of Statcast data is the metrics presented as average. I discussed this a bit last summer. It's time to expound on the problem with a focus on exit velocity.

Here are 2019's average exit velocity and BABIP of the three primary batted ball types.

Numbers vary from source to source, so it's important to maintain consistency within an investigation. That said, the exact velocity is secondary to the pattern. The exit velocity results are intuitive. Neither grounders nor flies are squared up, but more energy is transferred to non-centered flyballs since the path of the swing, usually uppercut, is in sync with the downward trajectory of the pitch. BABIP also makes sense, with the primary observation grounders have a higher BABIP than flies, despite a slower exit velocity.

The indirect correlation of exit velocity and BABIP is telling when looking at all-encompassing average exit velocity. Prevailing wisdom suggests the higher the average exit velocity, the higher the BABIP. However, this isn't always true. The correlation coefficient between BABIP and average exit velocity (minimum 200 plate appearances) last season was .11. A correlation of 1.0 indicates a perfect relationship while 0.0 means the relationship is random. There's some correlation, likely driven by line drives, but for the most part a higher exit velocity isn't always indicative of a higher BABIP. The ratio of groundballs to flyballs is integral.

Not all groundballs, flyballs or line drives are created equal. They can be struck with differing degrees of authority. Here's a table displaying BABIP of batted balls at the various levels of how well they were hit last season:

This is another instance of varying sources being an issue, as there was a time I used data showing the BABIP of medium-struck groundballs to be lower than those weakly hit. My rationale was a medium-hit grounder wasn't difficult to get to, but harder to leg out than a slow roller. There's a chance the classification of hard, medium and weak changed, or perhaps increased use of the shift changed the outcomes, with the reminder shifting isn't always effective.

Obviously, cutoffs are necessary and even though they're becoming more objective than subjective, a bunch of each hit type land near the boundaries, so exact numbers aren't as interesting as patterns. The data I want to focus on for the rest of this discussion is groundballs. Flyballs will be dissected in a follow-up study.

It was concluded earlier that average exit velocity doesn't tell the whole story; it's necessary to look at the components. The table above reveals a big advantage to a well-hit groundball. Let's run with that.

Circling back to the notion Statcast expected stats don't consider park factors, concentrating on just grounders alleviates that concern. Sure, a few venues still have turf and some infields play differently than others, but by and large, the area between the base paths is fairly uniform.

SPOILER ALERT: The ensuing analysis isn't nearly as informative as I hoped, but it's worth sharing, nonetheless.

Sparing the nerdy math, there's a slight correlation between average exit velocity on grounders and BABIP. A closer examination reveals the strongest correlation is at the top – the harder the grounder, the higher the BABIP. Thanks, Lord Obvious.

The next test involves trying to determine the reasons a player's groundball BABIP outperforms the expected mark as dictated by average exit velocity. Intuitively, speed seems like a clear-cut candidate. Sure enough, the correlation coefficient comparing the percentage difference from expected BABIP, and home to first time, is around -.40 (negative because the shorter the time, the better). Using sprint speed, it increases to .42.

There is a direct link between foot speed and groundball BABIP but it's smaller than I expected. I figured it would be higher. It's also a little surprising "time to first" had less of a correlation than sprint speed, though .02 isn't much difference. Since time to first accounts for which side of the box the batter started from, the assumption was it would be a better indicator. Perhaps batters don't go all out every time so that gets baked into the measurement.

Keeping in mind the ultimate reason for doing this was to first understand why a hitter posted their specific BABIP, with the hope of getting a glimpse of what the future holds, it seems there must be something more than running speed fueling a BABIP over what's expected based on exit velocity.

Lord Obvious wonders, what if foot speed doesn't have as big an influence on harder-hit grounders? Sure enough, the correlation between time to first and ability to outperform BABIP increases with lower exit velocities, but it's still just in the .50 range, so there's more to it still.

Let's look at some actual numbers. Below is the average ground ball exit velocity for every major leaguer with at least 50 grounders each of the past three seasons:

The league average was 86.3 mph last season, 86.1 in 2018 and 83.8 in 2017. In order to investigate further, the numbers need to be normalized so the total within each season is the same.

After normalization, the average and standard deviation for each player was determined. The higher the correlation between standard deviation and average, the likelier it is to maintain that level. This isn't a great test, but it's a quick way of learning something about the consistency. The results indicate there was a greater chance lower exit velocities would increase than higher ones decline. Again, not award-winning research, but good to know.

There are 226 players in the above sample. Last season, 118 (52.2%) improved their average exit velocity on grounders. The previous season, 112 (49.6%) hit grounders harder than the previous campaign. Essentially, it was 50/50, suggesting the chance of improving was random.

Looking at the results for consecutive seasons per player, if purely random, 25% would improve twice, 25% would decline twice with the remaining half up one season, down the next. The data shows 48 (21.2%) players improved twice, 134 (59.2%) improved once while 44 (19.4%) dipped twice. It's a tad off from purely random, but close enough to be considered sample size noise, so we really can't draw any tangible conclusions from a three-year spread.

The final exercise is displaying the above three groups to see if anything can be gleaned by knowing the names. Please keep in mind the numbers below are the normalized average exit velocities on grounders.

IMPROVED TWICE

IMPROVED ONCE, DECLINED ONCE

DECLINED TWICE

Keeping with the theme of much ado about nothing, there's a mix of young and old, speedy and plodding within each group. The final set appears to have more older players, but it also contains some of the best players in the game.

A change in exit velocity can result from several factors such as bat speed, swing angle and how close the contact came to centering the ball. Next week, a similar study will be done on flyballs. It will be interesting to contrast groundball and flyball results. If the player increased in both, the reason could they've been swinging harder -- or maybe they were making more consistent contact, which should manifest in fewer strikeouts. If the player's exit velocity on flies increased but dropped on grounders, that could be indicative of a swing change incorporating more loft. If the reverse is true, the swing could be flatter. This all speaks towards getting a better idea of what happened, so a more educated guess can be made for their performance this season.

Back when I was in graduate school, we were taught no result is a result. Saying there's nothing actionable from today's missive isn't true, but admittedly I was hoping for something more. With that in mind, here's a synopsis of what was learned (some rather intuitive) and the applications.

• A high average exit velocity increases the chance for a hit
• There isn't a big difference between a medium and weakly hit grounder, but foot speed is beneficial
• The ability to maintain groundball exit velocity gains looks to be random

Putting it all together, each player needs to be examined individually as opposed to being able to use specific filters to classify groups and identify expectations, which is the preferred goal of a study of this nature.

Thanks for hanging in. I have a feeling we'll learn more from next week's look at average exit velocity on flyballs.