Statcast 2021–2026 · 699,693 balls in play

T-Mobile Park, measured

Statcast’s expected stats are functions of exit velocity and launch angle, both fixed at contact. They have no term for where the ball went or how far it carried. Seattle is the park that exploits both hardest. Pick a ballpark and a contact cell to see what the model leaves out.

The blind spot

One xBA value, the entire width of the field

Every ball in a cell left the bat at nearly the same speed and angle, so the model assigns them all nearly the same value — the flat dashed line. The solid line is what happened, against spray angle.

xBA by exit velocity and launch angle

.00 1.00 paler = fewer balls

League-wide. Click any cell to drive the chart at right.

Actual outcome by spray angle

Every line is one ballpark, plotted against the direction the ball was hit. Green is the park you have selected; blue is every other ballpark, and clicking one selects it; grey is the league mean across all of them. The dashed rule is xBA, which is flat because direction is not one of its inputs. Green whiskers are the selected park’s 95% interval, and the bars along the baseline are how many league balls fall in each bin.

The same cell, on the field

The gradient

Where the outs live

Every batted ball placed where it came down, binned by direction and distance, coloured by how far the outcome fell short of or beat xBA.

Field

−.30 +.30 faded = interval spans zero

League average, same bins

Side by side, the park’s signature is the difference between these two fans, not the shape of either one.

The physics

Carry: the part defense cannot explain

Each ball hit in the air against the league’s median distance for its exact exit velocity and launch angle. Positioning cannot change how far a ball flies, so this isolates the park from its defence. Coors Field on top is the control.

Carry versus league expectation, feet

What actually goes missing

It does not eat home runs

Each hit type as a share of balls in play, against the league’s share of the same. Read the hit counts before the bars — triples are half a percent of all contact, so a huge percentage there is a handful of hits, while a small percentage on singles is many.

Hit types vs league

Shortfall by batted-ball type

BA − xBA relative to league, split by how the ball was struck. A carry effect should bite on balls hit in the air and leave ground balls alone.

The cost

How much of it is the building

The chart above counts hit types. This one prices them, which asks a different question: not what a park removes but what it cost. Because mean wOBA is a weighted sum of outcome rates, the shortfall splits exactly — every bar is that outcome’s own contribution, and they add to the total underneath. Splitting the same total the other way separates the part contact quality accounts for from the part it does not, and the second is the park. Left is every ballpark on wOBA − xwOBA; right is the selected one, taken apart.

wOBA − xwOBA relative to league

Coors Field and Sutter Health Park at the bottom are the control: both should be strongly positive and both are.

Where the wOBA went

wOBA points, against the league

The strikeouts

Contact never made

Personnel held fixed: each park scored once by following the home club’s hitters in and out of the building, once by following its pitchers, then averaged. 1.00 is neutral; 1.10 is ten percent more strikeouts than the same people manage elsewhere.

Strikeout park factor, personnel held fixed

The ledger

Every park, measured against its own contact quality

Measured against the league rather than zero — xBA runs slightly under reality on this population, so league average is a small positive number. The adjusted column divides out the hitters, charging each batted ball against that batter’s own record in every other park.

BA − xBA relative to league

Park table

Season by season

Is it stable?

A park effect that flips sign every year is noise. One that holds is a building.

Precision

What the sample will and will not support

Every chart above is an estimate, and they are not equally solid. Batting average is a coin flip repeated: its uncertainty falls as one over the square root of the count, so a measurement over 22,000 balls is roughly a fifth as wide as the same measurement over one season. Below, each headline number with the sample behind it and the width of its 95% interval.

Every estimate, with its interval

the last column is the estimate divided by its own margin of error

Below 1.0 in that column the interval spans zero and the measurement cannot tell you the direction, only that something is there.

What it takes to see an effect