Where does the same shot land?
One side-on court view with a physically proportioned court and net. Set the shot speed, topspin, contact, and launch angle to see the resulting trajectory.
Model takeaway: launch angle is the input; net clearance is the result. For a neutral rally ball, explore the angle that produces roughly 0.5–0.7 m of ball-center clearance, then notice how much depth margin remains. Lower clearances spend more of the error budget near the net; steeper launches pull the landing point toward the far baseline.
Physics model and validation · early-flight v1
Loading held-out model checks…
| Prediction error (RMSE) | Previous | New |
|---|
The new forecast observes only the first 20–180 ms to estimate position and velocity in 3D, then predicts the rest using gravity, drag, and signed spin lift. The previous replay used up to 380 ms to estimate launch angle. These tests do not establish accuracy from exact racket contact. The replay feed itself may be smoothed.
Only one tracked long flight passed all final-test requirements, so this is not a reliable test of long-shot-specific accuracy. This recalibration uses men’s matches and has not been independently validated on women’s matches. Wind and sidespin are not modeled; net height is held at the 0.914 m center height. In/net/long percentages use assumed execution variation, not measured personal error rates.
Download aggregate validation results · no raw tracking data is included.
Male match-shot distributions
The local 2026 first-round sample, split into forehand, backhand, and slice so unlike trajectories are not averaged together. These are quality-filtered groundstroke flights with a tracked bounce; percentages are population shares, not recommendations.
| Shot | Share | Speed (mean / median) | Spin magnitude (mean / median) | Contact distance from baseline (mean / median) | Net clearance (mean / median) |
|---|
Forehand means provider shot type 1 excluding negative-spin shots; backhand means shot type 6 excluding negative-spin shots; slice means negative-spin groundstrokes of either side. Slice rpm is shown as spin magnitude. Each distribution is a shared-scale small multiple with one row per shot type: bars show the bucketed shape, while the row header reports the mean and median. Contact distance from baseline uses a signed convention: positive values are behind the baseline and negative values are inside the court. Net-clearance observations require a tracked net crossing, so missed balls are not represented in that distribution.
Tracked long-flight distributions
Loading the tracked long-flight subset…
A tracked long flight is one whose tracked bounce is beyond the far baseline. The long-bounce population is complete at the event level, but contact and net-event packets are not present on every row, so each chart reports its own sample size. Speed and spin use provider records within 40–110 mph and 0–6,000 rpm; contact position and net clearance use the directly observed values that fall within the plotted court ranges. Mean and median are marked with dashed and dotted vertical lines, respectively.
How contact position and height shape the shot
Loading grouped means and medians…
Each point is a one-metre position bin or a 0.2-metre height bin; the lines are descriptive group summaries, not causal effects. Distance inside the hitter baseline is positive when contact is inside the baseline and negative when contact is behind it. “Short balls” means the tracked bounce was within ±1.0 m of the opponent service line.
Player shot replay
Choose a player and shot type to see 5 randomly selected tracked flights from that player’s local match sample. Solid lines are tracked paths; dashed lines forecast the rest of the flight from position and 3D velocity estimated during the first 180 ms, plus recorded spin. Speed colors use the provider’s launch-speed reading. Shots without a reliable early state keep their solid path but have no dashed forecast.
Loading the compact local shot sample…
The replay asset contains only a compact sample of tracked flights and shot telemetry; the raw tracking archive stays local. All 5 lines share a net-aligned side view so differences in contact distance from baseline, height, speed, spin, clearance, and landing depth are easy to compare. Genuine event-level flights that landed beyond the far baseline remain in the ordinary random pool when available and are marked LONG; they are not synthetic extrapolations. Contact distance from baseline uses a signed convention, not the service line.
Speed is launch speed in mph. The 4.5 and 5.0 markers are recreational estimates, not measured rating-level standards. The Pro reference comes from the earlier local sample of 42,981 quality-filtered positive-topspin groundstrokes: mode about 73 mph, 1,620 rpm and 1.05 m contact height; spin median about 1,920 rpm. Launch angle is the initial direction above horizontal. Manual shots travel straight along the court; replay forecasts retain the measured crosscourt component before projection into the side view. Contact position is relative to the hitter’s baseline, positive inside the court, from 3 m behind to 6.5 m inside. Service-line guides are about 5.49 m inside each baseline.
The solver uses gravity, quadratic drag and signed Magnus force, with no polynomial path correction. Net clearance is measured from the ball center to the tape; clearing it needs another 3.35 cm for the ball’s radius. The flight ends at tape contact or when the ball center is 3.35 cm above ground; no net-impact or bounce response is simulated. The suggested clearance range is an exploration starting point, not a proven optimum. Larger screens use equal horizontal and vertical metric scales.