SERVESPEED GUIDES
Video vs radar: comparing tennis serve speed results
Two numbers labelled “serve speed” may describe different parts of a serve, or come from different kinds of evidence. Before deciding which is closer to reality, check what each method measures. This is a guide to planning and interpreting a comparison, not a completed servespeed validation experiment or a claim that video estimates match radar.
Start with the meaning of each number
A visual estimate uses the available video evidence. A separate distance-over-time calculation produces an average over a defined interval. A radar device reports a value according to its measurement setup and operating mode. Sharing a unit, such as km/h, does not make these the same measurement.
| Approach | What the result describes | What to check |
|---|---|---|
| Video-based visual estimate | An estimate supported by visible motion and context; servespeed may return a range or no speed. | Ball visibility, camera perspective, timing and the evidence available. |
| Manual distance divided by time | Average speed over the particular path and time interval selected. | The physical distance, timeline and exact start and end events. |
| Radar reading | A device measurement that may report a detected peak or a live reading, depending on mode. | Alignment, model, settings and the part of the ball flight captured. |
An average is not a peak
OpenStax defines average speed as distance travelled divided by elapsed time. A calculation from contact to a later event describes that whole interval. It does not establish the highest speed within it or the speed at contact. A radar peak and a flight-time average can therefore differ without either having measured the same quantity incorrectly.
For a manual calculation, document the chosen events and distance. A straight line on the screen is not automatically the physical ball path. Perspective, unclear frames and altered playback speed can affect the inputs. This calculation is general physics background, not a description of how servespeed produces its visual estimate.
Read the radar mode and placement instructions
The Stalker Pro II+ operator manual distinguishes peak serve speed from live speed as the ball slows in Tennis mode. It also explains angle error: an uncompensated Doppler reading can be lower when the target travels at an angle to the radar rather than directly toward or away from it.
That example is specific to a documented device; other models have their own modes and procedures. Follow the manual for the device used and record its position and settings. Do not assume every displayed radar number is the speed at racket contact, or correct a reading using a guessed angle.
Compare paired records from the same serve
A useful comparison keeps the evidence for each serve together. Comparing a radar reading from one attempt with a video estimate from another introduces an uncontrolled difference before the methods are even considered.
- Give each serve a simple reference number. Keep the original video and the corresponding radar display or saved reading under that reference.
- Record camera position, recording settings and any trimming or export. Keep the recording at its original speed.
- Record the radar model, mode, units, placement and relevant settings. Note whether the value is a peak or a live reading.
- If calculating an average separately, record the start and end events, elapsed time, distance and how each input was obtained.
- Save the servespeed result exactly as shown, including a range or unavailable speed. Keep missing readings and unclear clips in the record rather than silently discarding them.
Investigate differences before drawing conclusions
First confirm the serve reference and units. Then check whether the methods describe comparable points or intervals, and review alignment, framing and timing. Do not use a fixed conversion factor to turn an arbitrary video estimate into a radar-equivalent number.
A servespeed range is an estimate, not a statistically validated confidence interval. A radar value falling inside it does not prove accuracy; a value outside it does not identify the cause. One matching pair cannot establish reliability across camera positions, devices or players.
Keep the observations separate from your interpretation. Note what the recording shows, which settings are known and what remains uncertain. A claim about accuracy needs a defined reference method, a documented procedure and suitable repeated measurements. No such comparison results are presented here.