The Safeco Field Effect: Batted Ball Velocity in 2015

In his last few seasons as General Manager, Jack Zduriencik attempted to build an offense that was focused on one thing: mashing the baseball. Based on the free agents he signed and the trades he made, it was clear that one of the main characteristics he was looking for in a hitter was power. Not just any kind of power either; he was specifically bringing in players who had a high average batted ball velocity off the bat. There are 13 players in the top 100 on the batted ball velocity leaderboard that played for the Zduriencik-led Mariners at some point. Nelson Cruz and Mark Trumbo are the most prominent examples of the type of player Zduriencik was obsessed with but even Justin Ruggiano, Mike Morse, and Jesus Montero made their way onto that leaderboard.

We’re all familiar with the unique circumstances that affect how baseball is played in Seattle. I’m talking about the suppressive marine layer present on most of the west coast. What that physically means is the air is denser at sea level. Objects traveling through denser air will need more velocity to "push" through the higher resistance. Zduriencik’s thinking was reasonable: succeeding in an environment that suppressed offensive output required players who could hit the ball hard to overcome their circumstances. At least, that was the theory.

Before this year, with the resources available to the public, proving this theory was completely out of reach. With the introduction of StatCast, more granular batted ball data is becoming available to the public, including batted ball velocity readings. There have been a number of studies that have shown the limitations of this data. Let me repeat myself: there are serious limitations to this data. Between the calibration errors and dropped data points, we can barely call StatCast data useful, but I think exploring the data is still a worthwhile endeavor. It’s there, it’s available, why not see what we can glean from it, even if it’s incomplete.

I’ll be using batting average on balls in play (BABIP) as our dependent variable. It’s a decent application of raw batted ball data; hitting the ball harder should result in more hits. It’s not perfect however: speed certainly plays an important factor in a player’s BABIP but I want to try and keep this as simple as possible (for my own sanity mostly).

First let’s examine the raw batted ball velocity readings from Safeco field:

	Avg. BB Velo. (mph)
League Avg.	88.4
Safeco Field	89.1
Mariners	89.9
Visitors	88.5

The league average batted ball velocity was 88.4 miles per hour and opposing teams maintained that average when visiting Safeco Field. Mariner batters hit the ball harder than league average at home and that’s a decent sign of Zduriencik’s plan in action.

Before we get into actual batted ball results, I have to explain a particular quirk in the data. About a quarter of the StatCast batted ball data does not include a velocity reading. And the batted balls without velocity reading heavily skew towards the odd and awkward hits that the equipment couldn’t get a reading on. That mean, on average, the data points with velocity readings are going to look much better than the data points without velocity readings. With that in mind, let’s see if BABIP tells us anything about the Safeco Field effect:

I’ve split the data into two pools, data with velocity readings and data without velocity readings. In the data pool with velocity readings, we can see that the Mariners’ BABIP was better than league average at home and much higher than their visiting opponents. This seems to show at least another point towards the success of the Zduriencik plan.

The data pool without velocity readings adds a bit of confusion to the data. Here, the visiting teams had the highest BABIP in the dataset and the Mariners’ BABIP is just above league average. Until we know specifically why these data points don’t have velocity readings, we can’t make any definitive claims as to why this is. For now, it’s just an odd quirk in the data.

Let’s dive into the performance of just the Mariners for a bit. I took the Mariners home and road batted ball velocity readings and split them into 10mph increments to see how the team performed across a range of outcomes. In the graphs below, you’ll see three lines, one for home performance (teal), performance on the road (blue), and the league average (grey). First, the Mariners’ BABIP across seven velocity bands:

(Edit: I noticed a data error in the previous version of this graph. The data has been corrected and the graph updated.)

When playing on the road, the Mariners’ BABIP was decidedly below average. At home, however, it reached league average in a few velocity bands. The most interesting group is balls hit between 81–90 miles per hour. These balls sit in what's known as the doughnut (balls that are hit well but not well enough to fall in for hits). The Mariners performed much better than league average on these kinds of balls in play; usually they're easy fly outs or liners that hang up too long in the gap and are chased down.

The highest velocity group is pretty disheartening from a team perspective. On balls that are hit the hardest, the Mariners performed well below average at home and very poorly on the road. The amount of home runs this team hits could be affecting the numbers a bit but it shouldn’t be that drastic and wouldn’t fully account for the brutal road BABIP.

What about the effect batted ball velocity had on run scoring? Let’s use wOBA as our measure of offensive quality and split the data between the same seven velocity bands, split by home and away:

Here we see an interesting trend: at lower velocities, the Mariners were more productive at home than league average. But at higher velocities, their production matched their road performance and fell below league average. Their success at lower velocities could be a benefit of Safeco Field. With the expansive outfield and the marine layer, bloops and dinks may be more likely to fall in for hits in Safeco Field. Since grounders are grouped together with fly balls and line drives in this dataset, that doesn’t tell us the full picture. And it doesn’t do anything to help us demonstrate the success of the Zduriencik plan.

Again, the Mariners’ performance at the highest velocity is very disappointing. When playing on the road, there was almost no discernable difference in production between a ball hit in the 101–110 range and a ball hit at speeds over 111 mph. That’s not a good thing.

These last two graphs might raise more questions than they answer. Why was the Mariners’ performance so bad on the road, especially on the hardest hit balls? If the goal of the Zduriencik plan was to build an offense that could succeed in Safeco, wouldn’t that success transfer to other stadiums with more favorable hitting environments? There’s honestly too much noise in the data to come to any firm conclusions. It may be as simple as playing better at home and enjoying a comfortable and familiar environment, no matter how oppressive the marine layer is.

Bringing in hitters who could mash the ball might have given the Mariners a slight advantage at home but it certainly didn’t create a killer offense that succeeded no matter where they played. From this cursory exploration of the data available to us we can see hints of an advantage, but when it comes to actual results, the Zduriencik plan just wasn’t able to succeed.