Skip to content

Tennis Status

A configuration for tennis serve return:

SettingValue
Response zonesWide / Body / T (3 zones)
Chance level33.3%
Occlusion anchors300 / 167 / 83 / 0 ms before contact
Starting offset83 ms
Range0 – 450 ms
Response deadline600 ms, tightening to 300 ms
Lead-in2000 ms

The wider offset range and earlier anchors relative to badminton reflect the longer approach of a serve action — there is more pre-contact kinematic information available, and it is available earlier.

The shorter deadline reflects the real task: a returner has less time to commit than a badminton receiver.

  • No tennis footage, and none validated
  • No sport picker in the app — it currently uses the first configured sport, so tennis is unreachable from the interface
  • No timing verification for tennis clip characteristics
  • No test–retest data establishing that the tennis configuration produces stable thresholds

Adding tennis must touch only configuration and assets — zero changes to the trial engine, the staircase, or the threshold estimator.

If any of those three require editing to support tennis, the sport abstraction has failed and must be fixed before proceeding. This is an explicit acceptance criterion, not an aspiration.

  1. A sport picker in the app — small piece of work
  2. Receiver-viewpoint serve footage — camera at the returner’s position behind the baseline, server’s full body and racket in frame at full extension, 120 fps. See Filming Protocol
  3. Contact frames marked and zones tagged as Wide/Body/T from the returner’s perspective
  4. A timing verification run on the tennis clips
  5. A test–retest pair exactly as required for badminton

Why serve return is a good first tennis application

Section titled “Why serve return is a good first tennis application”

The stroke is self-paced, so the stimulus onset is unambiguous and every clip has the same structure. There is no rally context to control for.

The viewpoint is unambiguous — the returner stands in one place, so “what the athlete would have seen” is well defined in a way it is not for open-play groundstrokes.

The three-way zone choice maps onto a real decision. A returner genuinely commits to wide, body, or T. The response is not an artefact of the test design.

Open-play groundstroke anticipation is a considerably harder measurement problem: variable court position, variable viewpoint, and a decision that is continuous rather than categorical.

These are worth settling before footage is shot, because they determine whether the test measures anything:

Is Wide / Body / T the right decomposition? It matches how returners talk about serve placement, but a deuce-court/ad-court split, or a first/second-serve distinction, might discriminate better.

Are the occlusion anchors right for a serve? The 300–0 ms range is a reasoned starting point, not an empirical one. The toss and trunk rotation begin much earlier than in a badminton stroke, so useful information may exist well before 300 ms.

Should the deadline be shorter? 600 → 300 ms is derived from realistic return times, but a returner’s actual commitment window on a first serve may be tighter still.

Input from a tennis coach on these three points is more valuable than any amount of engineering.