Cycling knee angles: interior angle, flexion and maximum extension
Understand which three points, angle convention and moment a number describes. Convert a fictional interior angle without mistaking it for a saddle-height instruction.
Published 8 October 2026 · OpenBikeFit
Camera field guide
Same directions. Different labels.
H = estimated hip · K = knee · A = ankle
Fictional: interior 145° → 180 − 145 = 35° flexion
Equivalent planar convention only; not a target or fit verdict.
On this page
Understand before recording
Inspect a fictional analysis
Use the existing free workflow and read its source and scope checks; opening a page does not accept consent or verify a rider.
Open toolA knee angle is incomplete without its definition. Two displays can show different numbers for the same drawn leg because one reports the interior angle between the thigh and lower leg and another reports flexion from a straight-leg reference. They can also describe different points, sides, frames or methods. Comparing the bare numbers before resolving these choices can create a disagreement that is only a labeling problem.
This guide explains conventions and the specific projected quantity used by OpenBikeFit. It is not a recommended knee angle or a saddle-height prescription. The worked values are fictional arithmetic. Do not move the saddle because an example happens to resemble a number in your report, and do not combine a static reference from one method with an unrelated camera result as though the measurements were interchangeable.
Start with three named points
For the knee, the app forms one direction from the estimated knee point toward the estimated hip and another toward the estimated ankle. Their included interior angle lies between zero and 180 degrees. A more open drawn knee has a larger interior angle; a more closed drawn knee has a smaller one. Those image points are model estimates, not physically attached markers verified against anatomical centres.
The actual side-view calculation scales normalized image coordinates by the frame width and height before forming these vectors. Treating both axes as the same length on a non-square image would change the geometry. This correct image aspect handling does not remove perspective, occlusion or landmark-placement error. It solves one mathematical convention, not every source of uncertainty in the picture.
Interior angle and flexion are different labels
In a fictional planar example, an interior angle of 145 degrees has a supplementary flexion value of 180 − 145 = 35 degrees. The two values describe the same two directions under those declared conventions; they are not competing fit targets. A straight drawn leg would have an interior angle of 180 degrees and a supplementary flexion value of zero.
Only make that conversion when the other measurement really uses the supplementary convention with equivalent segment directions. The formula cannot align different anatomical landmarks, compensate for a different camera view or change a moving observation into a static one. Before comparing reports, write the convention and method next to each number. If either definition is missing, keep the comparison unresolved.
Maximum extension is a cycle observation
The app extracts a series of knee interior angles while the rider pedals, conditions that series and segments cycles. Its knee extension summary uses the maximum interior-angle observation in each accepted cycle, then aggregates the cycle values. It is a maximum-extension estimate from the analyzed image series, not the angle of any arbitrary paused screenshot.
You may see bottom-dead-centre terminology around the knee metric in existing reports. That is an estimated extension-related metric: the side-view pose calculation does not directly measure the crank spindle or certify the exact mechanical instant when the crank is vertical. Maximum extension and crank bottom position should not be silently treated as identical measurements. Ankle posture and the timing of motion can matter.
Keep static and dynamic methods separate
A static reading is taken while a pose is held; a dynamic reading describes motion during pedalling. Different equipment may also use different landmarks. In a published comparison, Holliday and colleagues found method-dependent joint-angle differences and cautioned against interchanging the tested static instruments with their 3D motion-analysis method. That result is about those studied methods, not a validation or correction factor for this app.
The practical consequence here is modest: record whether the number came from a held pose, a moving clip, a manually drawn frame or an automated estimate. Do not add a universal offset to make it match a desired range. A convention conversion changes a label for equivalent geometry; a supposed accuracy correction would require evidence for the exact recording and measurement system.
A reference band is not a decision by itself
The existing app can show a mechanical reference band alongside its projected estimate. Its goal, context and evidence limits belong with that display. This article does not change the band, expand the supported scope or override capture, safety, repeatability and current-source gates. A number inside a band does not demonstrate that the rider is comfortable or that a component has been installed correctly.
Likewise, a number outside a band does not prove that the saddle is the cause or that moving it will produce a particular improvement. A camera-derived angle and a measured bicycle coordinate are different quantities. Review the exact metric and recording conditions first, and use a separate physical measurement for the bicycle. Hardware work still needs the exact component instructions and appropriate judgment.
Read a report without losing its context
Keep these items together: side, named points, interior or flexion convention, moving or static method, cycle summary, declared setup, quality band and date. An exported movement image has a narrower inventory than a complete report and must not become a detached approval to alter the bicycle. The complete report and the limitations should remain the reference for understanding its source.
Use the fictional public analysis to inspect labels before starting a camera workflow. To compare your own recorded dates, use the existing comparator only after its setup and protocol checks establish that the sources are eligible. Do not create a second record by copying a screenshot and call it independent evidence. The companion repeatability guide explains why repeated takes and before-and-after conditions answer different questions.
Practical questions
Frequently asked questions
Why does one report say 145° and another 35°?
They may use supplementary interior and flexion conventions: 180 − 145 = 35 for the same two planar segment directions. Confirm the points, side, method and timing before applying that arithmetic.
Is maximum knee extension exactly bottom dead centre?
Not necessarily. OpenBikeFit derives maximum extension from a conditioned projected angle series; it does not directly certify the crank-spindle position or exact vertical-crank instant.
What knee angle should I aim for?
This article defines a measured quantity rather than an individual target. A reference band must be interpreted with its method, goal, quality and scope; neither the fictional example nor a camera overlay is a fit verdict.