Guide
How to film your sprint for video analysis
The short answer
Film from roughly 10–15 metres away, perpendicular to the running lane, with the phone held steady at hip height and the highest frame rate your phone offers: 240fps if available, 60fps as a workable minimum. Capture the whole of the phase you care about with a couple of strides of margin either side, and keep the camera still rather than panning if the athlete will stay in frame.
Last updated
Where to stand
The classic advice is to stand perpendicular to the lane, and for most tools that is not advice but a requirement: they measure angles in the flat image, so any deviation from square distorts every joint angle they report. If your tool works in two dimensions, being off by fifteen degrees is enough to change the numbers meaningfully.
Distance matters as much as angle. Stand too close and the athlete leaves the frame before you have a full stride; stand too far and there are not enough pixels on the limbs for tracking to be reliable. Ten to fifteen metres is a reasonable default for a phone at 1× zoom, filming a single phase.
Frame rate, and why it decides what you can measure
Frame rate sets the finest time interval you can resolve. At 60fps each frame covers 16.7 milliseconds, so a ground contact of 0.09 seconds spans about five and a half frames. Meaning your measurement of it can be wrong by nearly twenty per cent from frame quantisation alone.
At 240fps each frame is 4.2 milliseconds and the same contact spans about twenty-two frames, which brings the error down to a few per cent. This is why research-grade contact-time work is done at 240fps and why any tool reporting contact time from 30fps footage should be treated sceptically.
Spatial measurements. Joint angles, touchdown distance, trunk lean. Are far less sensitive to frame rate. If you only care about positions rather than timing, 60fps is genuinely fine.
Lighting and background
Pose tracking needs to distinguish limbs from what is behind them. Bright, even daylight is ideal. Strong backlight. Filming into a low sun. Silhouettes the athlete and is the single most common cause of a clip that will not track.
A busy background with other athletes moving through the frame is survivable for good trackers and fatal for weak ones. If you have the choice, film when the lane behind is clear.
What to capture
Decide which phase you are analysing before you film, because the requirements differ. For a block start, frame the blocks and the first three or four contacts. For maximum velocity, film a flying section, the athlete should already be at speed when they enter frame.
Give yourself a stride or two of margin either side of the section you want. Analysis tools need complete strides, and a clip that starts mid-contact wastes the first one.
The five mistakes that ruin a clip
- Panning to follow the athlete. Moving the camera changes the reference frame mid-measurement.
- Filming into the sun, which silhouettes the limbs and defeats tracking.
- Standing too far away, leaving too few pixels on the joints.
- Using a default 30fps setting when the phone offers 120 or 240.
- Starting the recording late, so the clip opens mid-stride with no complete cycle.
Questions3
What frame rate do I need to film sprint analysis?
Use the highest your phone offers. At 240fps each frame lasts 4.2 milliseconds, which is what you need for accurate ground contact and flight time. 60fps, at 16.7 milliseconds per frame, is workable for joint angles and positions but introduces meaningful error into timing measurements. Avoid 30fps for anything time-based.
Do I have to film from the side?
For most analysis tools, yes, they measure angles in the two-dimensional image, so a camera that is not perpendicular to the runner reports distorted values. Tools that do not measure in the flat image, such as Stride, accept footage from any angle.
How far away should I stand to film a sprint?
Roughly 10 to 15 metres for a phone at 1× zoom filming a single phase. Closer and the athlete exits the frame before a full stride is captured; further and there are too few pixels on the limbs for reliable joint tracking.