A Draggy-style performance meter that runs on one iPhone — no external GNSS receiver. Put the phone in the car, hit START, and forget about it: the app detects the launch on its own and times everything at once.
Cruising at 55 mph and you floor it? The clock starts by itself as you cross 60, and from there it keeps timing 60–90, 60–125, 90–125, 125–130, 125–155 and so on. A run doesn't stop at a round number — as long as the car keeps pulling, the measurement continues.
| Standing start | 0–60, 0–100, 0–150, 0–200, 0–250 km/h (and mph equivalents) |
| Rolling start | Any pair of marks: 100–200, 150–200, 200–210 … 200–250 |
| Distances | 60 ft, 100 m, 1/8 mile, 1/4 mile, 1/2 mile, 1 km — with trap speed |
| Drag strip | Optional 1-foot rollout, trap speed over the last 66 ft |
| Classic | 60–130 mph in one line |
| Custom intervals | Any range, e.g. 130–170 — timed on every run |
| Chart | Speed curve; press and drag to read time, speed, distance and g at any point |
| Run conditions | Temperature, altitude, density altitude, track slope |
| Route map | The GPS track drawn on Apple Maps under every saved run |
| Background | Keeps measuring with the screen off or while you're in another app |
| History | Every run stored on device with its chart and full table |
| Units | km/h and mph, meters and feet, °C and °F |
| Appearance | Light, dark or system theme; eight accent colors |
GPS on an iPhone updates roughly once per second — nowhere near enough for hundredths of a second. So Dragly doesn't rely on GPS alone.
GPS (1 Hz, Doppler) ─┐
├─► Kalman filter ─► 100 Hz speed ─► interpolation ─► time
IMU (100 Hz, accel) ─┘ [v, bias]
- Kalman sensor fusion. The state is speed plus accelerometer bias. Prediction runs on every IMU tick (10 ms); correction runs on every GPS fix using Doppler speed weighted by its reported variance.
- Fix latency compensation. A fix arrives stamped in the past, so the innovation is computed against the estimate at fix time, not "now". Without this there was a systematic error of ≈0.16 s.
- Orientation doesn't matter. The phone can lie in any position: the direction of travel is learned automatically and protected against flipping under braking.
- Exact crossing times. Each mark crossing is linearly interpolated between filter ticks — 0.01 s resolution.
- Standing launches are caught on the acceleration edge (the IMU reacts an order of magnitude sooner than GPS), replaying the confirmation window so the first 0.2 s aren't lost.
- Slope and DA. Track slope comes from the barometer (≈0.1 m vertically versus meters from GPS); density altitude is derived from pressure and temperature.
- Works without the IMU too. If the accelerometer is unavailable the engine falls back to GPS-only — accuracy drops, but the run is still timed.
The accelerometer alone can be fooled — shake the phone hard enough and naive integration invents a launch. Three independent guards stop that:
- Longitudinal acceleration is clamped at 1.6 g. Even a slick-shod drag car stays below that, so anything higher isn't the car and never reaches the speed estimate.
- A standing launch requires 0.3 s of push in one direction — every tick must stay inside a cone around the first. Shaking reverses sign every few ticks and never confirms. Timing doesn't suffer: the run is stamped back to the start of the push.
- GPS has to corroborate the run. If Doppler speed never rises to at least half the fused estimate, the run isn't saved. Mid-run, a divergence above 6 m/s or 3 s without a usable GPS speed aborts it and re-anchors the filter on GPS.
The engine is run against synthetic physics (noisy GPS with delivery latency + biased accelerometer) and compared to an analytical reference:
| Scenario | Error |
|---|---|
| Standing start, 0–100 km/h | ±0.04 s |
| Rolling start, 100–200 km/h | ±0.02 s |
| 1/4 mile (ET) | ±0.01 s |
| GPS-only, no accelerometer | ±0.13 s |
| Worst case across 12 runs with different noise | ±0.16 s |
Real-world accuracy depends on GPS reception. The phone must be stationary relative to the car — in a mount, on a seat or in a pocket; held in your hand the IMU gets noisy and accuracy degrades toward the GPS-only level.
Download Dragly.ipa from the latest release. The build is unsigned — sign it with your own Apple ID using Sideloadly, AltStore or Xcode.
git clone https://github.com/ipadev88/Dragly.git
cd Dragly
open Dragly.xcodeprojBuild the Dragly scheme on your device (⌘R). Requires Xcode 26+ and iOS 26+.
Apple system frameworks only — no third-party dependencies, no SPM, no CocoaPods.
| Framework | Used for |
|---|---|
SwiftUI |
The entire interface |
Observation |
@Observable models instead of ObservableObject |
SwiftData |
Run history |
Charts |
Speed chart with scrubbing |
CoreLocation |
Doppler speed and coordinates |
CoreMotion |
CMDeviceMotion at 100 Hz + CMAltimeter |
Foundation |
The algorithm core — no UI or sensor dependencies |
Dragly/
├── Engine/ core, pure Foundation — testable off-device
│ ├── KalmanSpeedEstimator GPS + IMU fusion, fix latency compensation
│ ├── RunEngine run state machine, mark detection
│ └── RunTypes result models, interval resolution
├── Services/ sensor wrappers
│ ├── LocationService CLLocationManager → SpeedFix
│ ├── MotionService CMDeviceMotion → AccelTick
│ ├── BarometerService CMAltimeter → pressure and slope
│ └── SimulatedDriveService synthetic run (DEBUG only)
├── Models/RunRecord SwiftData model of a run
├── App/AppModel wiring: services → engine → store
└── Views/ screens: measure, result, history, settings
The core (Engine/) imports neither CoreLocation nor CoreMotion nor SwiftUI — only Foundation. That's why the algorithm compiles and can be verified with plain swiftc on a Mac, without a simulator and without a car.
| Permission | Why |
|---|---|
| Location (when in use) | Doppler speed is the basis of every measurement |
| Motion & fitness | Accelerometer and barometer for accuracy between GPS fixes |
Background location is enabled only while a measurement session is armed, so iOS shows its blue location indicator exactly when Dragly is actually measuring — and never otherwise. "Always" authorization is not requested. Continuous navigation-grade GPS plus a 100 Hz IMU do drain the battery, so press STOP when you're done.
Nothing is uploaded; your data stays on the device. The only network request is air temperature for the run's coordinates via Open-Meteo (no keys, no account). Offline the app works fully — it just won't show temperature.
The interface ships in English and Russian, following the system language.