Bike Fit Analyzer
Open the analyzer →
The four bike fit angles, and where the target ranges come from
A side-on fit read comes down to four angles. None of them is a score out of ten; each is a proxy for something specific that goes wrong when a contact point is in the wrong place. Here is what each one is actually measuring, the range it should land in, and — the part most articles skip — why the numbers you find online contradict each other.
The targets, in one table
| Angle | Target | What it is a proxy for |
|---|---|---|
| Knee flexion at the bottom of the stroke | 30–40° | Saddle height |
| Torso from horizontal | 40–50° | Overall aggressiveness of the position |
| Elbow bend | 15–30° | Whether your arms are absorbing road shock |
| Shoulder angle (torso to upper arm) | 80–95° | Reach — how open the cockpit is |
Those are the ranges the analyzer grades against, for a road-endurance position. The torso range in particular shifts with the bike: a TT position lives lower, a hybrid or commuter lives much more upright, and grading a city bike against road numbers would just paint everything red.
1. Knee flexion at the bottom of the stroke
This is the most useful number in the whole report, and the most reliable one to measure from video. It is the angle your knee makes at the bottom of the pedal stroke, expressed as flexion — how far short of straight your leg is. A completely straight leg is 0°. The target is 30–40°.
What it tells you is saddle height, and the logic is direct. Too much flexion at the bottom means your leg never gets to extend, which is a saddle that is too low: you lose power and load the front of the knee. Too little means your leg is reaching for the pedal, which is a saddle that is too high: your hips start rocking to follow it and the back of the knee takes the strain.
Roughly: above about 42° the saddle is likely low, below about 28° it is likely high. Between those the answer is that your saddle is fine.
One detail worth knowing: the analyzer measures this at the bottom of every revolution in your clip and takes the median, not the average. A median ignores outliers, so one frame where the pose model lost your ankle cannot drag the result. It also only grades this angle at the bottom of the stroke — mid-stroke your knee is bent far past 40° by design, and colouring that red would be meaningless.
2. Torso angle from horizontal
Measured from your hip to your shoulder, against level ground. 0° would be lying flat on the top tube; 90° would be sitting bolt upright. The road-endurance target is 40–50°.
This is the one angle where "correct" depends most on you. It is set by saddle-to-bar drop and by reach, but what makes a given drop appropriate is your hamstring and lower-back flexibility. A rider who can hold 38° comfortably is not doing anything wrong, and a rider forced into 38° by a bike that is too long for them will be uncomfortable at exactly the same number.
Above roughly 56° the position is quite upright — fine for a commuter, a lot of drag on a road bike. Below about 34° is genuinely aggressive and only sustainable with the flexibility to support it.
3. Elbow bend
Target 15–30°. The thing this really detects is locked-out arms. An elbow reading near 0° means your arms are straight, and straight arms transmit every bump directly into your shoulders and neck. Bent elbows are your suspension.
Locked arms usually mean you are reaching — the bars are far enough away that straightening is the only way to get to them. That is why elbow and shoulder tend to go wrong together, and why the analyzer colours them as a pair on the skeleton overlay.
The clinical reference here is a mean elbow bend of about 19° with a standard deviation of 8°, which is worth remembering when you see a fit article quote a single precise number: the population spread is wider than the differences people argue about.
4. Shoulder angle
The angle at your shoulder between your torso and your upper arm. Target 80–95°.
This is the reach proxy, and it is the least precise of the four. It depends on where exactly your hands are on the bars, which varies through a ride, and shoulder landmarks are the hardest for a pose model to place from a single side-on view. A reading well below 80° suggests a closed, scrunched cockpit with your weight sitting on your hands.
Treat this one gently. If your shoulder angle reads a bit low but nothing hurts, that is not a reason to buy a different stem. Reach is also the last thing to change, for reasons the adjustment order guide goes into.
Why published bike fit numbers disagree
Search for the correct knee angle and you will find 25–35° on some sites and 30–40° on others. Both are right, and the difference is not sloppiness — it is when the measurement is taken.
Static measurements are taken with a goniometer while the rider sits still with the pedal at the bottom. Dynamic measurements come from video while the rider is actually pedalling. The two do not agree: angles measured while pedalling run about 8° higher than static ones. Your ankle drops differently under load, your hips move, and you are not holding a pose.
So a static 25–35° and a dynamic 30–40° describe the same well-fitted rider. Video analysis is dynamic by definition, which is why this tool uses the higher band. If you have been measured by a fitter with a goniometer and get a different number here, that offset is most of the explanation.
How much precision is real
A phone video resolves joint positions to a few pixels, which works out to a few degrees of angle. The analyzer allows about 2.5° of tolerance at the edge of each band for exactly this reason, and it will not pretend to a precision it does not have.
The practical consequence: a reading of 41° against a 40° limit is not meaningfully different from 39°. Do not chase the last degree. Do pay attention when something is 10° outside its band, because that is well clear of the noise floor.
What about the front and rear views?
Knee tracking and hip rock are measured from the front and rear, and they are labelled beta and reported as hints rather than graded. That is deliberate: 2D frontal-plane measurement has a smallest detectable difference of roughly 8–9°, so any threshold tighter than that is inside the method's own noise. Healthy, correctly-fitted riders also show real frontal-plane movement — about 7° of pelvic rock is normal — so a tight threshold would flag most people who are perfectly fine.
See your own four numbers, graded against the ranges above.
Sources
- Holmes method and joint-angle averages — the clinical standard for knee angle at the bottom of the stroke (knee 36 ± 7°, elbow 19 ± 8°).
- Dynamic vs static validity study — the ~8° offset between pedalling and static measurement.
- Cycling kinematics in healthy adults — frontal-plane range of motion in correctly-fitted riders.
- Reliability of 2D frontal-plane knee assessment — the measurement noise floor.