SWING LABBASEBALL · SOFTBALLFOUNDER BETA · EXPERIMENTAL
Method
How this Founder Beta reads a side-view clip. Citations are the primary studies listed below.
Side-view lock
The only capture angle is the opposite batter's box, chest-facing, near hip height, roughly perpendicular. Rear, front, and catcher views fail closed. Angle tolerance is not validated.
Landmark set
Pose reading uses the GOLF_CORE_13 set: nose, shoulders, elbows, wrists, hips, knees, and ankles (indices 0, 11–16, 23–28). Heels and toes are not used. There is no bat landmark.
Gates
G5 checks side-on view. G6 trims a single swing window. G1–G3 require landmark coverage through that window. Failures return unavailable rather than a priority.
Own-baseline rules
Stride, hip shift, and head-path consistency flag against your own mean ± 2 SD once six swings exist on this device, or against a prior session's baseline, and only when both the in-sample and leave-one-out checks reach 2 SD. With three to five swings the lab shows averages only. Front-knee collapse flags when the lead knee still drops 3° or more over the last two frames at contact on at least two of the last three swings, and is ignored when the knee is already 175° or straighter. Knee-at-contact flags are low confidence when contact is estimated from motion. Head drop is a visual cue only and is never flagged. Status: calibration-v3, provisional.
Youth
A guardian should supervise. Adult published numbers are not applied as cut-offs. Activate, Discover, and Progress keep side-view measurements and do not get a mechanical priority. Published lab context is hidden for those bands. Height is optional, local only, and never turned into a personal percent-of-height figure.
Fortenbaugh 2011 (cited background only)
In a 3D lab study of 43 AA professional hitters, the front knee was about 15–18° bent at contact (Fortenbaugh 2011, University of Miami dissertation). That contact-knee figure was not tied to batted-ball outcome in that study (p=0.355), so this lab does not treat it as a quality check. A 2D estimate can be off by 7–22°, so this is background only. The lab judges brace by direction against your own swings.
In that lab average, the lead knee straightened by roughly 30° between the end of coiling and contact (about 48° bent down to 18°). That flex-then-straighten pattern is why the brace check looks for a positive direction. Escamilla (2009) measured about 59° of extension from maximum flexion in adults — a different window, not a contradiction. This page does not compare your degrees to 30.
Two pro lab studies put stride length around 40–50% of height (Fortenbaugh 2011; Welch 1995). This preview does not convert your clip into a personal percent-of-height figure. The 40–50% range is cited background only.
https://hittingperformancelab.com/wp-content/uploads/2015/01/hitting-baseball-miami-study.pdf
Primary studies
- Fortenbaugh DM (2011). The Biomechanics of the Baseball Swing. PhD dissertation, University of Miami.
- Escamilla RF et al. (2009). A comparison of age level on baseball hitting kinematics. J Appl Biomech 25(3):210–218.
- Welch CM, Banks SA, Cook FF, Draovitch P (1995). Hitting a baseball: a biomechanical description. JOSPT 22(5):193–201.
- Orishimo KF et al. (2023/2024). Lower extremity kinematic and kinetic factors associated with the baseball swing. Sports Biomechanics.
- Nakata H, Miura A, Yoshie M, Kudo K (2012). Differences in the head movement during baseball batting between skilled players and novices. J Strength Cond Res 26(10):2632–2640.
- Rode D et al. (2025). Assessment of monocular human pose estimation models for clinical movement analysis. Scientific Reports 15:38767. 2D knee flexion mean absolute error about 9–22°.