Last updated: May 2026 · Analysis engine: Google MediaPipe Pose

How AxyZ AI Bike Fit Analysis Works

AxyZ uses computer vision to extract biomechanical measurements from a short side-view video of you riding. This page explains exactly what the AI measures, how it works, and what the results mean for your fit.

The Core Technology: Pose Estimation

AxyZ analyzes video using MediaPipe Pose, a computer vision model developed by Google Research that detects 33 body landmarks in real time. The model tracks the position of your joints — including the hip, knee, ankle, shoulder, and elbow — in every frame of your video.

AxyZ samples at 10 frames per second regardless of the source video frame rate — enough temporal resolution to capture full pedal cycles while keeping processing fast. For a 10-second clip, this yields approximately 100 analyzed frames. It identifies the frames that correspond to the top and bottom of the pedal stroke (top dead center and bottom dead center) and extracts angle measurements at each of those positions across multiple pedal cycles. This multi-cycle averaging reduces the impact of frame noise and gives a more reliable measurement than any single snapshot.

MediaPipe Pose is the same computer vision foundation used in professional sports analysis applications, physical therapy tools, and motion capture research. AxyZ applies it specifically to cycling biomechanics using established bike fitting benchmarks as the reference standard for each metric.

The Four Metrics AxyZ Measures

1. Saddle Height — Knee Angle at Bottom Dead Center

Saddle height is the most impactful variable in bike fitting. AxyZ measures the angle at your knee joint when the pedaling crank is at the bottom of its rotation (bottom dead center, or BDC). The established benchmark for most cyclists is 35–40 degrees of knee flexion at BDC — equivalent to approximately 140–145 degrees of knee extension, consistent with the guidelines published by BikeDynamics and BikeFitAdviser, and with the biomechanical research of Holmes, Pruitt, and Whalen (1994) on lower extremity overuse in cycling.

A saddle set too low leaves the knee over-flexed at BDC (above 40 degrees), compressing the patella and causing anterior knee pain. A saddle set too high leaves the knee under-flexed (below 35 degrees), forcing the hip to rock laterally on each downstroke to reach the pedal — straining the IT band and reducing power transfer. AxyZ accounts for your crank length (which you enter before upload) when calculating the target range, since longer cranks shift the optimal saddle height.

2. Hip Angle — Hip Flexion at Top Dead Center

Hip angle is measured at the top of the pedal stroke (top dead center, or TDC), where the hip is at its most closed (maximum flexion). The target minimum hip angle for most cycling disciplines is 60–69 degrees at TDC — sufficient hip opening to generate force at the top of the stroke without impinging the hip flexors.

A hip angle below 60 degrees (too closed) indicates the saddle may be positioned too high relative to the handlebars, or the cockpit reach is too long, compressing the hip flexors on the upstroke. An angle above 69 degrees (too open) suggests the rider is sitting more upright than optimal for the discipline — common when the saddle is too low or the handlebar stack is too high. This metric is particularly relevant for triathletes and time-trial riders, where aerodynamic position must be balanced against hip clearance and hip flexor health.

3. Torso Lean — Upper Body Posture

Torso lean measures the angle of the rider's upper body relative to vertical. Unlike saddle height, there is no universal target for torso angle — the ideal position depends on the rider's discipline, flexibility, and comfort goals. Road cyclists targeting aerodynamics typically aim for 40–45 degrees of torso lean; recreational and fitness riders are typically comfortable at 55–70 degrees; mountain bikers vary significantly by terrain.

AxyZ scores torso lean against a range appropriate for the bike type you select during upload. It also notes whether shoulder position is symmetrical and whether the rider's weight appears to be over-loaded onto the hands — a sign of excessive reach or insufficient core engagement.

4. Pedaling Stability — Hip Drop Across the Pedal Stroke

Pedaling stability measures how much the hip drops vertically on the downstroke — the signature pattern of a saddle set too high. When the saddle exceeds the rider's effective leg length, the hip must dip to reach the pedal at bottom dead center. From a side-view video, this vertical hip movement is the most reliable measurable indicator of saddle height overshoot.

AxyZ tracks the hip landmark's vertical position across every sampled frame and calculates the variance across multiple pedal cycles. A stable fit shows minimal variance; a saddle that is too high produces a consistent periodic drop that the AI flags as a priority adjustment.

The X, Y, and Z Axes — What AxyZ Covers

In biomechanics and motion capture, body movement is described across three spatial axes. The name AxyZ reflects exactly which of those axes a side-view video can reliably measure — and which it cannot.

  • X axis — sagittal plane: Forward and backward motion: knee extension and flexion, hip angle, and torso lean. This is the primary plane of the pedal stroke and the source of most fit-related discomfort. All four of AxyZ's metrics are derived from sagittal-plane measurements captured in a side-view video.
  • Y axis — vertical: Up and down movement: the hip drop that occurs stroke-by-stroke when a saddle is set too high and the rider must reach down to meet the pedal. AxyZ tracks vertical hip position across multiple pedal cycles and scores the variance.
  • Z axis — frontal plane: True side-to-side movement: lateral knee tracking, cleat alignment, foot pronation and supination. A side-view camera cannot capture Z-axis motion — it requires a front or rear camera, or full 3D motion capture. This is the primary domain where a professional fitter with multi-camera setup adds value that a single-camera video analysis cannot replicate.

Focusing on X and Y is not a limitation of ambition — it is a deliberate accuracy boundary. Side-view 2D analysis produces reliable, actionable data for the sagittal and vertical planes. Reporting on Z-axis variables from a single side-view camera would introduce measurement error that could result in incorrect adjustments. AxyZ reports only what it can measure accurately.

How to Get the Most Accurate Results

AxyZ's accuracy depends primarily on camera placement and video quality. Follow these guidelines for the most reliable measurements:

  • Position your phone level with your bottom bracket — the center of the crank spindle.
  • Stand 6–10 feet away so your full body and the bike are in frame with room to spare.
  • Film from the right side so the drive chain is visible — AxyZ is calibrated for right-side analysis.
  • Record at least 10 seconds of steady-state riding at a comfortable cadence (not sprinting or climbing out of the saddle).
  • Use good lighting. Backlighting (bright window behind you) will reduce landmark detection accuracy.
  • Wear fitted clothing. Baggy clothing obscures joint landmarks and reduces precision.

Accuracy and Limitations

AxyZ is a directional tool. It identifies which fit variables are likely outside your optimal range and ranks the adjustments that will have the highest impact on your comfort and efficiency. It is not a replacement for a professional in-person bike fitting, which uses 3D motion capture, pressure mapping, flexibility assessment, and physical examination over 1–3 hours.

The primary limitation of 2D video analysis is that it captures only one plane of motion. Lateral knee tracking, cleat alignment, and fore-aft saddle position are difficult to assess reliably from a side-only view. AxyZ focuses on the four metrics where side-view 2D analysis produces reliable, actionable data — saddle height, hip angle, torso lean, and pelvic stability — and avoids reporting on variables where the measurement would be unreliable.

Crank length affects the geometry of the pedal stroke. AxyZ asks for your crank length during upload and uses it to adjust the target angle ranges for saddle height and hip angle. Cranks typically range from 165mm (shorter riders, more hip-clearance-focused) to 175mm (taller riders, traditional road geometry). Using the wrong crank length in the upload form will shift your scores, so measure the number printed on your crank arm before submitting.

What the Report Includes

After processing your video, AxyZ returns a scored report covering all four metrics. Each metric receives a numerical score and a status (optimal, minor adjustment needed, or significant adjustment needed). The report also includes:

  • A prioritized adjustment list, ranked by expected impact on comfort and power.
  • Specific, actionable recommendations — not generic advice. If your saddle is too low, the report estimates by how much.
  • Context for each recommendation, explaining the biomechanical reason for the adjustment.
  • Discomfort correlation — if you noted knee pain, back pain, or hand numbness during upload, the report maps those symptoms to the relevant metrics.

Pro users can save analyses to their fit history and compare results across sessions — useful for tracking the effect of saddle changes, weight shifts, or new bike setups over time.

Sources & References

Holmes JC, Pruitt AL, Whalen NJ. “Lower extremity overuse in bicycling.” Clinics in Sports Medicine, 13(1), 1994. — foundational study on knee flexion targets and injury correlation in cyclists.

BikeDynamics — UK-based professional bike fitting studio; published saddle height and knee angle guidelines used as reference benchmarks.

BikeFitAdviser — independent bike fitting resource; reference for hip angle and torso lean norms by discipline.

International Bike Fitting Institute (IBFI) — professional standards body for bike fitting certification and methodology.