2D vs 3D Running Gait Analysis: What Each Method Can and Cannot Measure

If you have looked into running gait analysis, you have run into a fork in the road: the high-end 3D motion-capture lab on one side, and a phone on a tripod on the other. They are not the same tool, and pretending otherwise does no one any favors. The honest version is that each measures a different slice of the same movement, and understanding that slice is what tells you which one you actually need. This piece lays out what 3D marker-based capture measures, what 2D single-camera video and markerless software can and cannot do reliably, and why a simple phone screening is still worth your time. The CritchPitch Run Lab is built on the 2D end of that spectrum, so being straight about its limits matters to us.

Gait analysis just means measuring how the body moves as you walk or run. The difference between the methods is how many dimensions they can see, and whether they can measure the forces underneath the motion at all.

The gold standard: 3D marker-based motion capture and force plates

The reference standard for human movement is a marker-based optoelectronic system. Reflective markers are placed on bony landmarks, and a ring of infrared cameras tracks each marker in space to reconstruct the position of every body segment. Because the system sees the body from many angles at once, it can compute joint angles in all three planes at the same time: the sagittal plane (flexion and extension, seen from the side), the frontal plane (adduction and abduction, seen from the front), and the transverse plane (internal and external rotation, the twist of a segment about its own long axis).[6]

The second half of the lab is the force plate. Embedded in the floor or built into a treadmill, it measures the ground reaction force, the push between foot and ground, in three directions. Combine those forces with the 3D motion and body-segment data, and the system can calculate joint moments, an estimate of the load and torque at the hip, knee, and ankle. This coupling of kinematics (motion) and kinetics (force) is what makes marker-based capture the benchmark that every other method is validated against.[7]

What 2D single-camera video measures

2D video analysis records the runner with one camera and reads angles and timing off that flat image. It is simple, portable, and cheap, which is why clinics have used it for years. A systematic review of 2D video gait analysis concluded the method can be reliable and valid for a number of parameters, while cautioning that performance varies widely by variable and depends heavily on how the video is captured.[1] The plane the camera faces decides what it can read.

Where 2D is strong: the sagittal plane, cadence, and timing

Filmed side-on, a single camera sits in the same plane as the biggest running motions, so sagittal-plane measures are its home turf. Studies of 2D running analysis report good repeatability for many sagittal variables such as knee flexion and trunk lean. Reinking and colleagues found intra-rater reliability for experienced raters ranging from 0.75 to 0.98 for sagittal measures, far tighter than their frontal-plane numbers.[2] Pipkin and colleagues similarly reported reproducible gait-event detection and substantial to excellent agreement for most kinematic ratings from 2D video.[4] Cadence and spatiotemporal timing, such as step timing and rhythm, are also well within reach, and a study in adolescent runners found 2D analysis reliable for nearly all kinematic variables except step width and foot progression.[3]

Where 2D is only an estimate: the frontal plane

Point the camera at the front or rear and things get shakier. A three-dimensional motion projected onto a flat image gets distorted, and small changes in camera angle shift the reading. The evidence is genuinely mixed. Maykut and colleagues found 2D and 3D methods correlated moderately for hip adduction (r roughly 0.54 to 0.62) but showed no significant correlation for contralateral pelvic drop, even though the 2D measures themselves were highly repeatable.[8] The lesson is that a frontal-plane number can be consistent from clip to clip yet still not match what a 3D system would report. Treat hip adduction, pelvic drop, and knee position from the front as directional estimates, not exact degrees.[1][8]

What 2D simply cannot do: rotation and force

Two things sit outside a single camera entirely. The first is true 3D rotation. Transverse-plane motion, such as internal rotation of the shin or the twist of the thigh about its long axis, cannot be recovered from one flat view, and it is the plane even 3D markerless systems struggle with most.[6] The second is force. Video shows motion, never the ground reaction forces or the loads inside a joint. Those require a force plate or an instrumented setup, full stop. Any claim that a plain phone video measures impact force or joint load is measuring something inferred, not something recorded.[7]

Markerless pose estimation: the accessible middle ground

Between the two poles sits markerless motion capture, where deep-learning pose estimation finds your joints in ordinary video with no markers at all. This is the technology moving fast, and the validation literature is refreshingly candid about the pattern. Kanko and colleagues compared a markerless system against marker-based capture during gait and found spatiotemporal parameters and lower-limb kinematics broadly similar to the gold standard, with agreement best in the sagittal plane and progressively weaker in the frontal and transverse planes.[5] Reviews of markerless systems echo this: agreement with marker-based capture is often excellent for sagittal angles and reasonable for the frontal plane, while long-axis rotations remain the least reliable output.[6]

Multi-camera markerless tools push further. OpenCap, an open-source platform, uses two or more smartphones plus musculoskeletal modeling to estimate 3D kinematics and even ground reaction forces, reporting joint-kinematic errors around 4 degrees against marker-based capture with equipment costing a fraction of a lab.[7] That is impressive, but note the setup: it takes multiple synchronized cameras and modeling assumptions to approach 3D force estimates. One phone, one angle, cannot get there.

2D vs 3D at a glance

So why bother with a 2D phone screening?

Because the gold standard is out of reach for almost everyone, and a good screening does not need to be the gold standard to be useful. The 3D lab answers research-grade questions about forces and rotation. A 2D screening answers a different, more practical one: are there movement patterns in how you run that are worth a closer look? For the measures that matter most to everyday runners, cadence, sagittal joint angles, over-striding, and step timing, 2D video is exactly the plane those show up in, and the reliability evidence supports reading them.[1][2][3]

Accessibility is the whole point. A lab session you cannot book teaches you nothing. A phone clip you can film this afternoon, and refilm next month to track a change, teaches you plenty, as long as you respect its limits and capture it cleanly. If you want to try it, our free tool lets you screen your stride from your own footage, and the guide on how to film your running gait walks through the camera angles and frame rates that keep a 2D screening honest.

The short version: 3D marker-based capture with force plates measures the most, in every plane, including the forces a camera can never see. 2D video measures a narrower slice, well for sagittal mechanics and timing, roughly for the frontal plane, and not at all for rotation or force. Knowing which slice you are looking at is what keeps a phone screening useful instead of overselling it. When you are ready to look at your own stride, you can screen your stride with that context in hand.