# Driving-feel evidence status (r344, 2026-10-16)

## Recovered and represented

- **Steering input:** The retail RealDriver registers digital and analog
  steering events. The gamepad preview samples the recovered 21-point
  controller curve; keyboard endpoints use the same `-1/+1` endpoints. The
  packed steering/Ackermann angle uses one full turn per 65,536 units, which
  the viewer preserves.
- **Throttle input:** Retail `0x00407a30` limits ordinary throttle-command
  changes to `+1.0/s` rising and `-1.5/s` falling. The viewer applies those
  bounds to keyboard/gamepad throttle before its approximate force model.
- **Update cadence:** Retail `PhysicsMover` splits frames at a maximum
  `0.0044 s` step. The viewer uses the same ceiling. Its road travel checks
  subdivide movement to at most `0.45 m` per check.
- **Player handoff:** Static code traces `RealDriver+0x1e0` into Ackermann
  `+0x90` through `PhysicsMover` when the relevant owner relationship holds.
  The mode and active object for the free-roam player have not been captured.

## Still preview behavior

The `0.22 rad/s` steering yaw cap, `3.5 m/s²` lateral estimate, direct heading
integration, `0.55 rad` wheel-angle display cap, `2.5 m/s²` acceleration
baseline, brake/coast/e-brake rates, and road-edge speed damping are viewer
choices. The source does not provide a measured link from the active driver's
steering word through tire forces to player yaw. It also does not provide
usable force units for engine output, braking, or per-car tire grip. GLOBALB
mass/engine/gear/wheel data help with relative previews and RPM, but they do
not determine absolute acceleration or lateral grip.

The retail event table has separate steering, throttle, brake, reverse, and
e-brake events. Default keyboard bindings and the active player drivetrain
consumption of brake/reverse/e-brake remain unverified. Therefore the viewer's
brake-to-reverse transition, brake deceleration, handbrake pivot, and
130 mph ceiling must not be described as recovered handling values. The
130 mph ceiling is the requested preview limit; it coincides with the 240SX's
source gearing at the 6500 RPM upshift target but is not proof of retail
maximum speed.

## Packaged observations

- [Route 736 start in the center marking](../webgl-city/review/images/491-r341-route736-centerline-before.png) and [after using the authored selector-4 corridor](../webgl-city/review/images/493-r341-packaged-route736-lane.png). The packaged page reached ready state, reported the 1.8056 m start offset, and had no page error. This is a narrow route-point placement correction, not a general lane-choice reconstruction.
- [Compressed combined-world drive](../webgl-city/review/images/499-r343-compressed-combined-l4rb-drive.png) shows the player car moving on the L4RB raised road after throttle input. It verifies movement and road support, not retail acceleration or steering calibration.
- [W+D curb run](../webgl-city/review/images/487-r340-combined-live-drive.png) shows the player at the road edge after sustained steering. Speed falls because `movePlayerAcrossRoad()` rejects unsupported positions and repeatedly damps speed. This is viewer boundary behavior; no retail curb response has been measured.

The next source-grounded handling step still requires a free-roam XP capture
correlating the active player controller, RealDriver command values, Ackermann
wheel transforms, four wheel response vectors, and RigidBody velocity/yaw over
matched input sweeps. See [`driving-feel-boundary-r308.md`](driving-feel-boundary-r308.md)
and [`vehicle-physics-runtime.md`](vehicle-physics-runtime.md).