Skip to content

v0.5.71 Results insights — source and installed-game QA

Player-facing source work

The genuine Unreal Results menu is extended with three independent, presentation-only systems:

  1. Recent timed-lap statistics analyze the last 2–8 retained, actively timed, valid completed laps. Average, median, within-window minimum/maximum, population standard deviation and signed newest versus window mean are computed. A strict decreasing-time streak counts only consecutive adjacent physical lap comparisons. Once the fixed-size eight-lap ring wraps, these metrics describe only the remaining recent window, not the full race history.
  2. Best-gate split review displays the actual best positive time for each scoring gate within the run, together with the mixed-lap theoretical sum. Best splits may come from different physical laps; do not label the sum an achieved lap time or a new record.
  3. Sampled car telemetry displays chassis speed-integrated distance (estimated, not an odometer), observed peak km/h, braking share of accepted time, brake activations and high lateral-slip sample time. It is not authoritative for official elapsed time or a skill grade.

The existing bounded Results pager shows these detail rows on desktop and Android. The compact view suppresses optional secondary lines rather than resizing the right-side touch pads or moving the warning band. Nothing here alters physics, opponents, checkpoint scoring, recovery fees, career rewards, settings, save schema, networking or persistent records.

Strict data-admission rules

  • Both timing reviews require CanAwardFinishedRace: missing active timing, zero-time and penalty-only laps, incomplete finishes and record-invalid clocks must not produce premium results.
  • Lap stats require two or more valid recent samples with contiguous physical lap IDs, newest equal to completed laps, and an exact window of min(completed laps, eight). No average silently incorporates laps already evicted from the bounded history buffer.
  • Standard deviation is the population deviation over the retained sample; the signed newest-minus-average comparison treats positive values as slower. Equal adjacent laps break the faster-lap streak.
  • The gate review requires finite positive per-gate observations and the exact theoretical best already computed by real RaceProgress. The source theoretical sum must agree and be no greater than the best achieved lap.
  • The sampled driving review requires at least two accepted frames and >=0.25s accepted sample time. Counters, distances, speeds, sample-time shares and physics-derived fractions are bounded and checked for finiteness. Missing evidence displays no review.

Portable regression tests

cmake -S Tests/Standalone -B Tests/Standalone/build -DCMAKE_BUILD_TYPE=Release
cmake --build Tests/Standalone/build --config Release
ctest --test-dir Tests/Standalone/build -C Release --output-on-failure
python3 -m unittest discover -s Tests/Editor -p '*Tests.py' -v

New strict C++20 targets compile actual RaceProgress.cpp or DrivingSessionStats.cpp rather than writing a simulation of their production state. Cases include three- and twelve-lap windows, sample median/mean/deviation, faster/slower/equal streaks, true mixed-lap gate bests, reset, missing/penalty-only/zero-time results, touch-screen compact display binding, interrupted chassis samples, numerical bounds and time-share calculations.

Unreal and installed-game acceptance (outstanding)

  1. Use a licensed Unreal Engine 5.8 installation; run UHT and build both editor and game targets. Source test CI is not this step.
  2. Author/cook the real original circuit asset and test a three-lap event. Deliberately achieve one best sector on a slower completed lap and confirm the theoretical mix is never represented as an achieved record or reward threshold.
  3. Verify recent-lap average, median, range and improvement streak against the actual engine timers. Repeat across a long (>8 lap) qualified event; older evicted laps must no longer influence stats.
  4. Inspect paused and finished Results in landscape/portrait Android, narrow Win64 desktop sizes, high DPI and non-ASCII car names. The UP/DOWN detail pager, touch buttons and SAVE FAILED message must remain unobstructed.
  5. Verify accumulated chassis telemetry during braking, cornering, normal pause/resume, recovery and restart. Understand speed- integrated distance is approximate and may omit invalid samples.
  6. Retest changing input focus, menu category, player possession, physical package launches and save reload; no late/ghost data may survive explicit restart or wrong-run state.
  7. Record actual editor compiler logs, build SHA, map/hash, device names/OS, screenshots, frame time, failures and package artifacts. Add physical Android proof to qualification #13.

No licensed engine test, cook, Win64 or Android game package, or physical device session is claimed by the source suite. v0.5.71 is a published source-only pre-release with zero executable binaries, not a playable or stable release.