A full biomechanics lab buys you things a camera genuinely cannot: ground reaction forces, true three dimensional joint kinetics, elbow torque in newton meters. It also costs a building, a staff, and a scheduled visit per athlete. This page draws the honest line, measurement by measurement, so you can tell which questions need a lab and which ones do not.
The term covers three different setups that get discussed as if they were one thing, and they answer different questions at very different cost.
| Setup | What it adds | What it costs you in practice |
|---|---|---|
| Marker based 3D capture with force plates | The reference standard. True 3D joint angles and, critically, kinetics: forces and torques, including elbow varus torque and ground reaction force. | A dedicated space, calibrated camera rigs, marker suits, trained staff, and a scheduled session per athlete. Athletes come to it. |
| Markerless 3D capture, multiple cameras | Most of the 3D kinematics without the marker suit. Increasingly used in stadiums and academies. | Fixed multi camera install and calibration. Still a place, not a workflow. |
| Single camera markerless | Sequencing, phase timing, and 2D joint angles from the film a program already shoots. | A phone. No kinetics, and every spatial number is a 2D approximation. |
The important thing about that table is the third column. Labs are not just expensive, they are a different unit of work: an appointment for one athlete, not a screen of a whole roster during a normal bullpen.
This is the table we would want to see from a vendor, so here it is about our own engine. Anything marked as needing a lab is something we do not report, at any confidence.
| Measurement | Needs a lab? | Why |
|---|---|---|
| Kinetic sequence order | No | The order the pelvis, trunk, elbow, and hand reach top speed is visible from one camera, and it is the most predictive thing in a delivery. |
| Phase timing | No | Load, drive, release, deceleration, and finish are detectable to the frame. Load to release duration and foot contact timing come out directly. |
| Joint angles | Partly | Real flexions read well from a side view as 2D approximations. Rotations toward or away from the camera do not, and we omit them rather than estimate. |
| Hip to shoulder separation | Partly | A single camera reads roughly 60% to 75% of the true 3D value, so we report it as a proxy with that stated. A lab gives you the real number. |
| Lead knee extension | No | Extension between foot contact and release is the single most valuable measurable here, worth roughly 1 mph per degree (Dowling et al. 2024), and it reads from video. |
| Elbow varus torque | Yes | A torque in newton meters requires force measurement and validated 3D kinematics. No camera produces this. Anyone reporting it from a phone clip is inventing it. |
| Ground reaction force | Yes | Needs force plates. There is no video proxy worth trusting. |
| Energy flow and joint power | Yes | Derived from kinetics, so it inherits the force plate requirement. |
| Release point consistency | Needs tracking | A ball tracking system, not a lab. From video you can judge mechanical repeatability across pitches instead. |
| Arm health and workload | Neither | Not answerable from any single session, in a lab or on video. We make no claims here and no responsible system should. |
If you are asking how much stress a joint is under, what the ground is giving back, or how energy transfers between segments in absolute terms, that is a kinetics question and it needs force measurement. A research program, a return to throw decision informed by clinical staff, or a validation study all belong in a lab.
If you are asking whether this pitcher sequences correctly, whether the front leg braces, whether separation peaks in the right window, whether the finish is controlled, or whether any of it changed since March, that is a pattern question. Patterns read from film, and they read on every athlete you already film instead of the few you can schedule.
Most day to day development questions are pattern questions. That is the whole reason a single camera engine is useful despite everything in the middle column above.
Tell us what you are trying to decide. If it needs a lab we will say so, and we will tell you what a lab will and will not settle.
hello@ballparkanalytics.comThe six checkpoints, the reference window for each, and the drills.
What an organization gets: roster screening, reports, and data access.
The four mechanical leaks that cost the most velocity.