R&D — Design | The Undercut Wiki

The two Design tabs commission projects: this season’s new parts on tab 2, next season’s regulation car on tab 5. Both spend the R&D pool (the flask), both use the same continuous size slider, and both produce a blueprint — never a physical part. What separates them is the baseline they roll from and what happens if you miss a deadline.

How to open it

Car icon → Design {year} (tab 2) or Design {year+1} (tab 5).
The tabs are labelled with the season’s real calendar year and the one after it.
Both are locked during pre-season (current_phase < 6).
Tab 5 stays closed until the Federation publishes next season’s regulations — see below.

The project-size slider

One draggable slider, position t ∈ [0, 1], shared by both Design tabs (design_size, default 0.4). It replaced the old Low/Medium/Big buttons; cost and gain interpolate continuously across the same endpoints.

Quantity	t = 0	t = 1	Formula
Total R&D WU	80	240	`80 + t × 160`
Workload / GP	20	70	`20 + t × 50`
Base perf gain	+0.005	+0.020	`0.005 + t × 0.015`

The readout beside it shows Tier · N WU · ~G GP @ 100% workload, where G = total_WU ÷ rnd_capacity rounded up. The tier label (Low under 0.34, Medium under 0.74, else Big) is cosmetic plus a stored bucket — the real cost is the continuous total_wu persisted on the project.

Historic trap, now fixed: accrual used to cap at the legacy tier bucket, which froze any project sized above it (a 121-WU “Low” stuck at 80/121 forever) and rolled every over-bucket carry-over straight into a rushed Poor part. Accrual now caps at the project’s own total_wu.

Design {year} — this season’s parts

Develop a New Part

Every component except the engine gets a row:

Column	Meaning
Part	Icon + name
Perf	The currently fitted part’s `current_perf` on the 0–100 scale
Expected	A live preview at the current slider size, with the signed delta
Action	`Design` — or `in progress` / `no WU`

The preview and the commit use the same function, so what you see is what you get.

dept_scale   = 0.5 + dept_level × 0.05                       (lv1 0.55 … lv10 1.00)
gain         = base_gain(t) × dept_scale
initial_perf = min(fitted.current_perf + gain, 1.10)
potential    = min(initial_perf + 0.03 + norm(rnd_level) × 0.12, 1.18)
fiability    = clamp(0.85 + norm(eng) × 0.15 + norm(rnd) × 0.10, 0.85, 1.15)
health       = 100

norm(level) = (level − 1) / 9.

Two things follow from this:

An in-season design evolves the current car. It baselines on the fitted part’s current_perf, so a developed part yields a developed successor.
Head-room is a margin, not an absolute ceiling. potential = initial_perf + 0.03 + R&D margin guarantees a real development arc no matter how developed the current part is. The old absolute formula produced “new” designs that arrived already maxed.

The engine is not designable. It is rolled once per season per motorist into a shared vintage; every customer of that manufacturer receives the identical unit.

Fix a Problem — trait remediation

Above the new-part table sits the trait-fix list: one row per discovered negative trait on the component’s default design, with a problem badge, a green ✓ Solved result badge, the WU cost and an estimated GP count.

Full mechanics live in R&D — Projects. The short version: 40 WU for a Yellow trait, 120 WU for a Red one, and the fix resolves the trait outright rather than downgrading it.

Clicking Design (or Start on a fix) opens a modal:

Task name and total WU cost.
Work orders / GP: [−] N [+] with free X / cap Y.
A live Time: ~G GP readout (cost ÷ allocation).
A “pool after launch” bar showing allocated (gold) vs total (green).
Launch project / Cancel.

The default allocation is all free capacity; the value is clamped to at least 1 and never above the project’s total cost (allocating 200 WU/GP to an 80-WU project is pointless). The allocation is then locked out of your R&D capacity until the project finishes or is cancelled.

Constraints on launching: one project per component at a time, and rnd_free_capacity > 0.

Delivery and the quality roll

Projects accrue wu_per_turn at every GP — and at every pre-season turn (advance_rnd_projects runs there too). When wu_invested ≥ total_wu:

The component’s version counter increments.
A quality tier is rolled — this is the surprise, and it happens at delivery, not at commission.
apply_quality_to_part applies the tier’s budget with a random perf/head-room split.
roll_part_traits rolls a fresh trait set for the new design.
The result is registered as a KnownDesign (a blueprint) — no physical part is created.
An email arrives with the new part’s card, compared against the current default spec.

The quality roll

level = 0.6 × responsible_dept + 0.4 × rnd_dept          (design_quality_level)
q     = (level − 1) / 9
roll  = quality_infra_bias × q + (1 − bias) × luck + roll_bonus

Shipped tuning: quality_infra_bias = 0.40, quality_roll_bonus = 0.0, quality_budget_multiplier = 1.0. Thresholds:

roll	Tier	Budget	Positive traits capped at
< 0.12	Poor	−0.010	none
< 0.42	Normal	0.000	none
< 0.74	Good	+0.016	Bronze
< 0.92	Very Good	+0.030	Silver
else	Perfect	+0.046	Gold

Luck dominates (60% of the roll at the shipped bias). A level-1 shop can land a Perfect part; a level-10 shop can land a Poor one. Infrastructure shifts the distribution, not the outcome.

The budget is split perf_bonus = budget × s, head_bonus = budget × (1 − s) with s ∈ [0.25, 0.75] — so identical projects can arrive “ready now, low ceiling” or “raw, big ceiling”. initial_perf caps at 1.12, potential at 1.18. A guaranteed head-room floor scales with tier (Poor +0.004 … Perfect +0.026).

Negative traits can roll at any tier. See Car Traits.

Design {year+1} — the next-season planner

It opens at the regulation break

Until the Federation publishes the technical regulations, the tab shows “Next-season planner not open yet” and no component table at all. The package is announced in one bundle at the first season break — after GP 5 (REGULATION_GPS = [5]). Showing every part as designable before any rules existed was a real bug; the planner is now deliberately closed.

The regulation package

One severity is rolled per season and expanded into the affected components, all mandatory:

Severity	Scope	Odds	Break WU bonus
None	nothing	6%	×1
Small	1–2 random parts	24%	×1
Medium	every part of 1 department	22%	×1.5
Heavy	every part of 2 departments	30%	×2
Full	every part, supplier-rolled included	18%	×3

Departments are scoped to Aero / Chassis / Handling / Engine. The bonus is a one-off top-up of the Engineering pool on the break GP, scaled so a Full reset is survivable in one season.

The banner summarises N mandatory — must redesign │ M optional — carry-over allowed, and turns red with a count when mandatory parts still have no project. The engine is excluded from that count — the supplier re-specs it automatically.

Row statuses

Badge	Meaning
`MANDATORY` (red)	Regulated, no project — homologation will be refused
`Mandatory ✓`	Regulated, project running
`Optional` (amber)	No rule change — carry over, or develop for a bonus
`Optional ✓`	Optional part with a project running
`Ready ✓` (green)	Project complete, spec locked, delivers at the rollover

The engine row shows REGULATED → Supplier re-specs automatically or Supplier → Current engine carries over.

An N+1 design is a FRESH roll

This is the rule most players get wrong. The projected stats come from next_year_design_baseline — the infrastructure/supplier base spec — not from the currently fitted, fully developed part.

baseline = infra_to_car_model(infrastructure, suppliers) → that component

A next-year design starts around the infra baseline plus the slider gain, with a full development arc ahead of it. This season’s development does not carry into next season’s concept. The Plan column shows a signed delta against your current part, and it is routinely negative — that is correct and expected: you regain it (and more) through next season’s development.

All four commissioning sites use this same baseline — the player’s tab, the headless auto-pilot, and both AI planners.

The optional-development bonus

A part with no rule change may simply carry over its year-N spec. If you spend WU on it anyway, it receives a “specific development” bonus of NON_MANDATORY_N1_DEV_BONUS = +0.004 on top of the slider gain — the incentive to develop an already-legal part.

Completion and carry-over

A project that completes mid-season is acknowledged immediately: its allocation is freed back to the pool (a done design must not keep reserving WU/GP), an email says the design is ready, and the Projects tab marks it ✓ Complete · Ready next season. The blueprint itself still lands at the rollover.
A project that is unfinished at the rollover is carried over as a current-season project (with wu_per_turn floored at 1) and keeps developing through the pre-season turns. No WU are lost. The rollover email says the redesign is “still in progress”.

The rushed Poor part

A mandatory regulated component (the auto-re-rolling engine excepted) with no completed next-year design at the rollover is force-homologated as a rushed Poor part.

Field	Rushed Poor value
`current_perf` = `initial_perf`	base `initial_perf` − 0.010, clamped `[0.80, 1.12]`
`potential`	= `current_perf` → zero head-room; it cannot be Improved at all
`quality`	Poor
`traits`	none — a botched design carries no engineered traits
`health`	100 (new, just bad)
version	base version + 1

The flow:

At the rollover, the botched blueprint is registered as a KnownDesign and the component is flagged in rushed_regulation_parts. The part is not fitted yet and no email is sent — the old spec stays on the cars.
Pre-season turns keep ticking R&D, so a carried-over redesign can still finish.
At the CarReview package build, the inventory is stripped and the newest known design of every component is manufactured and fitted. If the real redesign finished during pre-season, its higher version wins and the Poor blueprint is never built. Otherwise the Poor part is the newest and goes on the car.
The email — “Rushed regulation designs — unfinished” — fires only for components whose newest design is still the zero-head-room stopgap.
The carry-over keeps developing in-season and delivers the proper part later at a still-higher version, replacing the stopgap.

AI teams take the identical penalty.

No comparison card for mandatory redesigns. A regulation redesign’s “design ready” email shows the new part’s card alone — the old spec is illegal under the new rules, so a “current → new” comparison would be meaningless. The comparison returns after the package build, when it measures the gain over the fitted Poor stopgap.

Choosing the default design

Producing, replacing and trait-fixing all read the component’s default design: the highest known version, unless you override it with Choose default in the Production tab. See R&D — Production.

A trait fix bumps the version so the fixed spec automatically becomes the default — unless you had pinned an older one, in which case your pin survives and the fix is never produced.

Advanced

Size the project against your capacity, not your ambition. ~G GP @ 100% workload assumes you dedicate the entire R&D pool to it. Two Medium projects at half allocation each take twice as long as the readout suggests.
Design early in the season, improve late. A delivered design starts near the infra baseline and needs Improve WU to reach its potential. Commissioning a Big project at GP 14 gives you a raw part and no time to develop it.
The R&D department is the head-room department. potential = initial_perf + 0.03 + norm(rnd) × 0.12 — R&D level 1 buys +0.03 of head-room (≈19 pts), level 10 buys +0.15 (≈94 pts). It also carries 40% of the quality roll’s infra weight. It is the single most leveraged department for a developing team.
Sequence the N+1 commissions. With a Heavy or Full package you cannot afford full allocation on everything. Commission the expensive/impactful parts first (body > floor > gearbox > … > fuel lines) — the AI does exactly this — and accept the Poor penalty on the cheapest ones.
A rushed Poor part is not just slow, it is frozen. Zero head-room means Improve shows Max forever. Only a fresh design restores potential.
The engine re-roll is regulation-gated. The shared vintage is only re-rolled when the engine is regulated that season; otherwise the fitted spec is frozen and carried forward at full life. Switching motorist during a freeze hands you the new brand’s pooled spec with a fresh, undiscovered trait set.
Pre-season turns count as GPs for R&D. A project sized to finish “in 3 GP” can complete across the pre-season instead. This is what saves most carry-overs from the rushed Poor.

Common mistakes

Expecting an N+1 design to inherit this year’s development. It never does — it is a fresh roll from the infra baseline. The negative delta in the Plan column is the intended behaviour.
Leaving a mandatory part unprojected “because there is time”. The deadline is the season rollover, and an unfinished project only saves you if it finishes during pre-season.
Sizing a Big project on a level-1 department. dept_scale = 0.55 cuts the gain nearly in half. 240 WU for +0.011 is a bad trade; raise the department first.
Treating a completed design as a part. It is a blueprint. Go to Production.
Reading the tier label as the cost. “Low” at t = 0.33 still costs 133 WU, not 80. Read the WU number.
Cancelling an N+1 project to free capacity late in the season. The invested WU are lost and you inherit the rushed Poor penalty on a mandatory part.