Each report runs the same engine: tracked release profile, arsenal shapes, an observed-vs-Magnus SSW read, and optimization targets as confidence bands. Snapshot arms populate metrics from the pipeline at the end-of-season refresh.
Built from Hawk-Eye / Statcast tracking. Release consistency is the command signal we watch; arm angle and extension set the plane and the perceived velocity. Kinematic-sequence depth (hip-shoulder sep, lead-leg block) is a Phase-2 add once force-plate or KinaTrax data is on file.
HB sign convention: positive = glove-side (catcher's view), negative = arm-side. IVB = induced vertical break vs. gravity. Categorization smoothed across Statcast/TM tags — Skenes' mix blends faster than the buckets.
Every pitch gets the same read: current shape, the SSW signature inferred from tracking, and an optimization target stated as a range with a confidence band — not a single number we can't actually promise off Hawk-Eye.
The old report quoted seam coordinates (e.g. "shift to 20H/10V"). Honest problem: Hawk-Eye can't see seam orientation — only a TrackMan unit with 3D Spin can. So here we infer the SSW signature from what tracking does measure: the gap between the break the spin should produce and the break the ball actually shows.
Active spin (spin efficiency) is the share of spin doing Magnus work. The rest is gyro or seam-effect. So per pitch:
SSW signature ≈ observed break − (active-spin × Magnus prediction)
A pitch with high observed movement but moderate efficiency carries an SSW/gyro share — that's the bar shown on each pitch. The splitter and sinker light up exactly where the original report said the seams were doing the work, which is the cross-check.
When 3D Spin is available for an arm, this section swaps to measured seam orientation and the inference caveat drops. Flag the data source per arm and the math upgrades itself.