Pit Strategy Plans | The Undercut Wiki

Before each race the game builds up to three strategy candidates for every car, then hands them to you as A / B / C in the Race Run Plan screen. The AI executes the chosen plan’s stops automatically, with real-world flexibility for traffic, safety cars and tyre state.

This page covers how the candidates are generated, why the letters are not fixed compound sequences, when each shape wins, and how the AI deviates from the plan under pressure.

The letters are a RANKING, not a recipe

This is the single most misread thing about the screen. generate_strategies builds its candidates by shape:

Candidate	Stops	Compound sequence
One-stop, soft first	1	Soft → Hard
One-stop, hard first	1	Hard → Soft
Two-stop (conditional)	2	built only when it is competitive

It then sorts them by estimated race time and renames them A, B, C in that order. So A is simply whichever plan the estimator rates fastest on this circuit, with this car, at this fuel load — it is not “the aggressive one”, and it is not always the two-stopper. On a low-degradation track A is routinely a one-stopper, and the two-stop candidate may not be offered at all.

Read the compound sequence printed on each card, never the letter.

When the two-stop is offered at all

The two-stop candidate only survives if its estimated race time lands within TWO_STOP_MAX_DEFICIT_S = 8 seconds of the best one-stop. Below that bar it is dropped, because the extra pit loss (roughly a full stop’s worth, ~20-25 s) is not recoverable by fresh-tyre pace on most layouts. It also needs enough race left to be meaningful — a short race never generates one.

That 8-second window is the interesting part strategically: a two-stop shown as slightly slower on paper is often the better race, because the estimate assumes clean air and no safety car. See Tire Compounds for the wear curve that drives the estimate.

Stint length clamps

Each candidate’s stint lengths are clamped to what the compound can actually sustain on this circuit (the estimator uses the same cliff curve the physics applies, so a Soft stint is never planned past its cliff). A plan you never see was usually killed by that clamp, not by the time sort.

Fuel

Each plan bakes a fuel target per stint: enough to reach the next stop (or the flag) plus a safety margin. The fuel figure you set in the Run Plan screen overrides the plan’s starting load — and the compound + fuel snapshot you confirm beats the strategy index if the two ever disagree. See Run Plans.

When a two-stop wins

Short pit lane (Short Pit Lane track trait, when traits land) — pit loss < 18s
Tire-shredder calendar — Soft cliff hits hard; can’t reach the flag on a single Soft stint
You’re chasing from midfield — pace gain on fresh tires lets you cut through traffic
Soft compound is way faster than Hard — high-grip soft, low-grip hard
Safety Car gambits — a well-timed SC during a 2-stop covers one of the pit losses

A two-stop with a poorly-rated tyre supplier is suicide: the Soft cliff arrives early, you stop again to swap to Hard, and you lose more than the one-stop would have.

When the one-stop, soft first (Soft → Hard) wins

The default for most weekends. Optimal when:

Pit loss is normal (12–18s)
Tire wear is moderate
You’re starting around your race-pace position (no aggressive recovery needed)
The field around you is on similar plans → strategy variance doesn’t help

This is the “no thinking required” shape, and on most weekends the estimator ranks it first — which is why it so often carries the letter A.

When the one-stop, hard first (Hard → Soft) wins

You’re starting up the grid (P3+ in a midfield car) — extending stint 1 protects the position
Hot conditions — Soft would overheat in stint 1; Hard handles it
You’re protecting a points finish late in the season — risk-averse run-in
Low-CON driver — Hard’s wider window covers driver inconsistency in stint 1
Long opening stint is favoured by track layout (many corners → Hard manages wear)

A defensive shape. It rarely wins races but reliably finishes them — and because the letters follow the time sort, it usually shows up as B or C rather than A.

Generating the plans

The strategy generator (sim/src/strategy.rs) computes:

Total race laps based on circuit length × race_distance_factor
Stop laps for each candidate — the two-stopper splits the race in thirds, the one-stoppers split it around half, and every stint is then clamped to what the compound can sustain on this track (the same cliff curve the physics uses).
Estimated lap time per phase using a tire-wear curve and fuel weight model — feeds into the EV evaluator
Fuel target at each stop = actual_fuel_per_lap × stint_laps_to_next_stop × 1.08 (8% safety margin)

Each candidate is scored by estimated total race time. That score does two things: it drops the two-stopper unless it lands within 8 s of the best one-stop, and it decides the A/B/C order — the letters are assigned AFTER the sort, so they describe ranking, not compounds.

What happens during the race

The AI executes the plan but is allowed to deviate in three cases:

Undercut window

If a rival pits early, the AI can pit you 1–2 laps later than planned to undercut on fresh tires.

Overcut window

If your tires are still in window and the rival ahead pits, the AI can extend your stint to hold position via track gap.

Emergency

Three triggers:

Fuel emergency — only fires if fuel_kg < laps_remaining × fpl × 1.05 (you genuinely can’t reach the flag). No more “buffer pit” with one lap to go.
Tire emergency — wear > 85% AND laps remaining > 3
Crash damage — limp-to-pit terminates the strategy; race is over for that driver

Safety Car window

If an SC deploys in your pit window, the AI can pit “free” (other cars are at SC pace, your pit-loss is much lower). But the AI now refuses an SC pit if you’re in points, tires can finish, and a rival is within 18s behind — you’d lose position for nothing. Back-of-grid cars still take the free SC pit.

Manual override

You can override the AI by clicking the Pit button in the race UI. It commits the car to pit at the next opportunity. Your own button beats the AI’s plan; the AI won’t undo a manual pit request.

Use the manual override for:

Reacting to a tire delamination warning
Forcing an undercut the AI didn’t see
Bringing the car in early on safety-car deployment
Pitting a damaged car you want to retire on your terms

Don’t use it to second-guess the AI on routine pit windows — the EV evaluator is quite good at spotting the right lap.

Reading the strategy in race UI

The pit strategy panel shows:

Active plan — A / B / C label
Next planned stop — lap number
Pit window — coloured band on the lap counter (target ± 2)
Stops remaining
Compound queued for the next stop
Fuel target for the next stop

When you cross into the pit window, the panel turns yellow. After the pit, it updates with the next stop (or “Final stint” if it was the last).

Strategy + driver match

Driver profile	Best strategy
High CON, smart pace management (INT 15+)	B or A — they handle stint transitions cleanly
Low CON, error-prone	C — Hard first reduces stint-1 mistake risk
Aggressive (AGG 18+)	A — they’ll gain on the chase laps after each stop
Conservative (AGG 5–8)	B — won’t capitalise on A’s recovery laps

Quick reference

Scenario	Pick
First career, mid-grid start	Whatever the estimator ranked A — read its compounds, then commit
Top-3 grid, defend the position	Hard first: a long opening stint protects track position
P10+ grid, podium ambition	The two-stop if it was offered — its 8 s deficit is often recovered in traffic
Hot calendar, normal driver	Hard first (Soft would overheat in stint 1)
Cold calendar	Soft first (it warms in stint 1)
Tyre-shredder calendar	Two-stop if offered, else hard first — avoid a long Soft stint
Protecting a points finish late in the season	One-stop, whichever the estimator rates faster
Short race	One-stop — a two-stop is rarely even generated

Ignore the letter, read the compound sequence. See Tire Compounds, Run Plans and Track Traits.