Build 25868291

build

Neuralic Race Engineer · brach, public · first seen

PICS live
2
Branches
1
Depots changed
—
Download delta
19:05 UTC
Time · UTC
Patch notes · published by the developer

Update 1.0.6 — From Laps to Setup 10/11/26

On Steam

h1]Update 1.0.6 — From Laps to Setup

Two messages reached us in different words. A review on Steam: you drive the clean laps NRE asks for, then twice as many, and still no setup arrives. And a driver on Assetto Corsa Competizione: NRE recognises the car and the track, but the track map stays at 0 of 3 valid laps, coverage at 0%, and no setup is ever proposed. Both were right.

This update is about the road from your laps to a setup you can load — on both simulators, on every circuit, for every car. The fixes come from real logs, setup files and telemetry from a driver's PC.

🏁 Assetto Corsa Competizione: the track map, the first setup, the full setup

  • FIXED: track mapping stuck at 0/3 valid laps and 0% coverage, however many laps you drove. To map a circuit NRE needs the distance driven along the lap, in metres. ACC only sends the position as a fraction of the lap, and its track-length field is documented as unused. In 1.0.5 the metres came only from a reference lap imported by hand. NRE now measures the lap length from ACC's own data, on clean laps only (no pit lane, no restart); until the first clean lap it records nothing rather than guess. Confirmed in a real session at Brands Hatch, and checked in tests on 25 ACC circuits, the 25 km Nürburgring 24h layout included.
  • FIXED: the first setup of a car could never appear. ACC creates a car's setup folder the first time you save a setup for it. Until then, NRE looked in the folder of a similar car: the 488 GT3 for the 296 GT3, the GT3 Evo for the GT3, the GT4 for the GT3 of the same brand. If that folder held an NRE_Optimized, NRE answered "already there" and created nothing for your car; if it held one of your setups, that setup became the base. NRE now uses only the folder of the car and the track you are driving.
  • FIXED: a Min or Max setup saved on purpose to explore the ranges became the base of NRE_Optimized — TC 0, ABS 0, wing 0, minimum pressures. Such saves are now recognised (nearly every setting at the same end of its range) and skipped: the base is the last real setup of the car. An NRE_Optimized built from one is rebuilt, and a copy of the old file is kept outside the game's folder.
  • FIXED: the full setup after 8 laps never started in ACC if you had accepted the AI notice in Le Mans Ultimate. The acknowledgement was stored per simulator, so in ACC it counted as never given and the AI features stayed off. It now applies to the whole installation, and acknowledgements already given are found.
  • FIXED: NRE called you into the pits before the full setup. During the 8 laps that open it, a pressures-only proposal said "PROPOSAL READY - press APPLY, pit, and LOAD 'NRE_Optimized'", and pitting starts the count again. In one afternoon at Brands Hatch the stints were 7, 2 and 6 laps long, so the full setup never started and only the pressures ever changed. The pressures now wait: the banner and the log say not to pit and how many laps are missing, and at lap 8 the pressures arrive with the full setup, in one stop. APPLY still works, and if you pit earlier anyway the pressures are offered at the box. Le Mans Ultimate, which counts the same 8 laps, follows the same rule.
  • FIXED: new tyre pressures started from a fixed table for the car class, not from the pressures you were running: a front-right tyre 0.1 PSI above the hot target was lowered by 0.9 PSI. NRE now starts from the cold pressures of the setup you left the pits with and adds what is missing to the hot target, at most 1.5 PSI per step. The class table is used only when those pressures cannot be read. The full setup follows the same rule.
  • FIXED: the log said "1 change PROPOSED" when no proposal existed, and "GATE 1 rejected" when the full setup had not even started; 40 ACC messages appeared as internal codes. They now say what actually happened, in six languages.
  • FIXED: the Telemetry Center opened in ACC looked for the setup loaded in Le Mans Ultimate, and copied an LMU setup into ACC's data folder.
  • NEW: the I/M/O temperatures you read in ACC's pit screen now calculate something. Before, entering them only said "within" or "beyond the limit" for each tyre. Now NRE works out, for each axle, the camber click they ask for, with the 9 °C front and 5 °C rear limits: inner shoulder too hot, one click less negative; outer shoulder hotter, one click more negative. The reading is kept, the click goes into the next package if the knowledge base confirms it, and the reading at your next pit stop judges it: improved, unchanged, worse or overcorrected. A worse click is undone. Camber is written in the setup file's own clicks, inside the range seen for that car.
  • FIXED: ACC does not publish the shoulder temperatures, and the camber estimate used the left-right temperature difference in their place. It no longer does: on ACC, camber comes only from your pit readings.
  • FIXED: data of Le Mans Ultimate showed up in Assetto Corsa Competizione, and the other way round. In ACC, NRE asked Le Mans Ultimate which car you were driving and got the last car you had driven there: in one real session a Ferrari 296 GT3 was given the limits of a Le Mans Ultimate Hypercar, and its package history was filed under that car's name. The Telemetry Center stayed on the data of the simulator you first opened it in, and reference laps and the selected session carried over when you switched. Each simulator now uses only its own data: the Telemetry Center, the Coach and the Ghost show only the sessions of the active one, and a history filed under the wrong name goes back to the right car.
  • NEW: proposals in the Telemetry Center for live sessions, on both simulators. It analysed only the wheels of imported files, and NRE recorded none while you drove, so on a live session it always found nothing. It now records the four wheels too (pressures, brakes, suspension; ACC does not publish the shoulders), and a new card shows the engine's package, what was deferred and why, and the proposal waiting for APPLY.
  • CHANGED: "Collected data: 0 records — 67 sessions" looked like a fault. The 67 were the laps analysed, and records come only from completed A/B experiments. The line now says what it counts: "Laps analysed … samples from A/B experiments …".

🗺️ Track maps, on both simulators

  • FIXED: driving two circuits without closing NRE — Brands Hatch and then the Hungaroring, or Monza and then Le Mans — the second one never got its map. The best lap belonged to the session, not to the track, and no Hungaroring lap could beat a Brands Hatch lap. Each track now keeps its own best lap, and the Ghost no longer shows the best lap of another track.
  • FIXED: a piece of a lap was recorded as a lap — NRE started mid-lap, ESC back to the pits, a change of track. That piece could become the track's map, and then the reference every real lap was compared against and rejected by. On one driver's PC the Le Mans map covered 1.6 km of a 13.6 km lap. Only complete laps count now, and a saved map that does not cover the lap is replaced by the first complete one, whatever its lap time.
  • FIXED: a lap longer than about 12 km was cut short before the line: Le Mans in Le Mans Ultimate (one lap was recorded as 12,590 m between two of 13,610 m), and the Nürburgring 24h in ACC. The limit now follows the length of the lap the game declares.
  • FIXED: in ACC the Nürburgring 24h layout shared its identity with the GP circuit, so the laps of one were rejected against the map of the other.

🧩 Setup packages: the whole car, measured, written, checked

  • NEW: changes now come as a coordinated package per stint, decided by the knowledge base on the car's balance measured corner by corner from your laps: speed, lateral acceleration and steering. The old estimate came from a simplified model that treated the steering input as an angle and saturated on half of the samples. In tests with a real mid-corner understeer it reported oversteer, and proposed the front anti-roll bar change that makes understeer worse.
  • NEW: no fixed number of changes. A package is the best combination of everything your laps justify on the whole car — springs, bars, dampers, bump stops, ride heights, aero, differential, brakes — on both simulators and in every stint. The complete analysis (porpoising, suspension travel and speed, traction-control interventions, lock-ups, weak braking, session type) feeds it too, and the knowledge base confirms or vetoes each of its changes. Each change must earn its place: one that brings less than it costs in a harder-to-read verification is left out as unnecessary. One mechanism per symptom, two independent ones if the symptom is severe; the others wait for the next stint.
  • NEW: every change shows its evidence, its confidence and its risk. What the evidence does not support is deferred, not applied, and the log says what would be needed. Ride heights and wings, which move the aero platform, change only on a direct measurement.
  • NEW: every package is written, then read back from the file and checked, before NRE tells you the setup is ready. NRE then checks from the game that you loaded it, or asks you to confirm after leaving the pits. A new package is never stacked on top of one you have not loaded.
  • NEW: the next stint verifies the package: verified, partial, no improvement, worse or not verifiable, with the reason. It also looks for side effects: a new or worse problem is traced to the change the knowledge base says can cause it. A change that made the car worse is the first thing undone in the next package.
  • FIXED: a change tried without benefit came back. In a real afternoon at Brands Hatch the same two changes returned two stints after being judged ineffective, stacked on the first ones. A direction already judged is now not proposed again while the setting stays where that test left it.
  • FIXED: "SETUP COMPLETE" appeared while a written package was still waiting to be loaded at the pits. Complete now comes only after the verdict.
  • NEW: Back to the previous setup (two clicks, so it cannot happen by accident), and Cancel the package before it is written.
  • NEW: when there is nothing to change, NRE says why: not enough data, contradictory measurements, or no setup problem found. "Not enough data" is no longer reported as "no problem".
  • ACC races: the package waits for your pit stop and is written there. A change the car does not allow (for example the fixed splitter of the Nissan GT-R) is never written, and front and rear changes that share one field in the ACC file no longer overwrite each other.
  • FIXED: one balance package per stint is the rule. Tyre pressures and brake ducts no longer count as "change 1 of 5", and a balance proposal no longer disappears a lap later under a pressures-only one.

🔧 Le Mans Ultimate

  • NEW: a full setup from the first stint. The first time you drive a car on a track, after 8 laps (2 to warm up, 6 analysed) the first package already covers the whole car, as described above, including the session type's objectives. Moves based on the simplified model's understeer and oversteer stay out. Camber is never changed on track.
  • FIXED: wet tyres on a dry setup (LMP2, GT3, LMP3). Pressing DRIVE could rewrite current.svm with the game API's compound number — and in these classes the API's 1, "Medium", is the setup file's 1, "Wet". NRE no longer writes the compound from the API. It follows the compound actually fitted, as the game reports it, translated with the names the game writes in your own setups for that car ("Medium", "Media", "100% Wet"…), and it repairs the files written by the old code. When the translation is not certain, no file is touched.
  • FIXED: camber and ride-height settings were capped at 20 clicks. On a BMW M4 LMGT3 (camber 0–40), a two-click step from 35 became 20. No fixed ceiling is left.
  • FIXED: online, the game's garage summary sometimes followed other cars — about 25 different ones in a single session, while the shared memory named yours for 584 laps out of 585 — and NRE rebuilt current.svm with another competitor's car. It now writes only when both sources name the same car.
  • FIXED: Long Beach (and, still hidden, Paul Ricard) could have two setup folders, one of them created by a third-party setup installer. NRE picked the one the game does not use, so the setup was never created. NRE now uses the game's folder.
  • FIXED: an entry whose name does not include the model — an Oreca 07 LMP2 entered under its team name — was treated as a GT3. NRE now uses the class the game declares.
  • NEW: the setup you loaded, on screen: its name, whether it has unsaved changes, and every adjustable setting with its value as the garage writes it (51.2:48.8, 6 (Balanced), -2.20 °). In the garage it follows your changes; on track it is the setup you left the garage with. On track the panel never shows what you were looking at in the garage as if it were in use.
  • NEW: every live lap knows its setup. In the Telemetry Center a Setup column says which setup each lap was driven with, and the session lists the setups you used, with what changed from one to the next (80% → 70%). A lap gets no setup rather than a wrong one: after a garage change that could not be confirmed, or when the game's summary describes another car.
  • The "setup ready" message now says how Le Mans Ultimate loads it: Load Setup in the garage, file NRE_2_USE_THIS_*. A proposal is never called a ready setup.

🔬 Telemetry Center: one study per session (Le Mans Ultimate)

  • NEW: the REPAIR candidate used to come from one pair of laps: change the pair, and it changed or vanished. There is now one session study, built from all valid laps and the same whichever pair you look at, and REPAIR prepares that one.
  • NEW: when every proposal is "observation only", NRE offers an exploratory study — the strongest proposal that passes the car, value, axle and range checks — and says why the evidence is weak. It is never applied on its own: you decide by comparing laps. In Assisted and Controlled Auto the study is prepared by itself, only for your latest outing; approving and applying stay yours.
  • FIXED: the analysis stayed the one made the first time a session was opened. It is now redone when the laps change, and proposals it no longer makes are marked as superseded, never deleted.
  • FIXED: rules for the rear axle were written on the front-left wheel — "rear fast bump −1" changed the front left from 5 to 4 and left the rears alone — and the same affected springs, ride heights, camber and pressures. Also fixed: "Corner ?" in the verdicts, and per-corner statistics that mixed different cars.

⚖️ Balance of Performance 1.4.2

  • NEW: BoP 1.4.2 (22 September 2026), imported from the official document: 714 entries, each read twice by two independent methods that agree field by field. It replaces 1.4 as the active version. Your BoP file is updated at start-up, with a backup, and versions you imported yourself are left alone. Entries of 1.4 that could not be verified are kept as history only.
  • The BoP is only a constraint: it never proposes a change. Wing and ride-height limits (LMGT3) are applied only when the version is confirmed for the game build you have installed. Otherwise the panel says the limits are waiting for verification, and a change that would need them is left out of the package.

🧪 Quality

  • 4,567 automated tests pass, plus 65 that click through the real pages in a real browser. New among them: 25 ACC circuits driven in one session, the first setup of every ACC car NRE knows, similar cars told apart, the real engine end to end on eight car and track combinations, two Brands Hatch afternoons rebuilt from their logs (the pressures, and the package that came back), whole-car packages on both simulators, replayed on real Le Mans Ultimate laps, a pit reading followed all the way to a camber click written in the ACC setup file, the switch from one simulator to the other with nothing carried across, and the setup loaded in Le Mans Ultimate followed from the garage to each lap.

A note on what changed and what did not



Nothing was removed. The changes go where the setup is decided, written and checked: a lap counts only if it is complete, a setup is ready only after it has been read back from the file, and a change is proposed only on a measurement.

The ACC track map has been confirmed in a real session. The rest has been verified with automated tests and with real files from real sessions, but not yet with the games running live. If something behaves differently once real telemetry is flowing, the diagnostic report button in the health check writes a file with everything needed to track it down.

Per-depot manifest changes

Depot changes

All depots
DepotNew manifestPrevious manifest
466605133939084806359855367799882327983144350