Icy tires → which force calculation gets used?
Imagine blending two tire-force ingredients. The line shows how strongly this code weights the ice-specific ingredient. The ordinary ingredient gets the remaining weight in this part of the calculation.
The material selector uses the ice-family curve if any wheel-material entry is Ice (3), Snow (21), or RoadIce (74); otherwise it uses the other-material curve. This is a weight inside one tire-force calculation, not the percentage of total grip, icing gained, or speed gained. The graph uses the model's piecewise-linear control points. It does not show a universal minimum icing percentage for a sustainable slide; that also depends on speed, angle, and other forces.
How icing itself rises or fades
These lines show a single tire's icing percentage over time if it stays in one condition. This is a different quantity from graph 1's force weight.
Idealized fixed conditions at driving speed above the tested model's threshold: Ice/RoadIce builds from 0 to 100% in 1,650 ms; grounded non-ice such as plastic loses 100 percentage points in 3,300 ms; air loses them in 6,000 ms. During decay the minimum is the smaller of starting icing and wetness at the transition. This graph holds wetness fixed; in play, wetness and contact can change.
Does right mode get stored before takeoff?
Example: switch raw input from full left to full right. The internal steering moves gradually. Drag the deadline to the last physics update with at least one wheel touching.
The right-mode test is strictly greater than +10%, on a physics update with any wheel in contact. This is a normalized internal steering value, not VehicleState's FLSteerAngle or raw InputSteer. If the threshold is not crossed before takeoff, the prior left mode stays. The 0.2 change per update was measured in the tested ice conditions; six affected updates means roughly 50–60 ms after the input switch, depending on update alignment. The first wheel to lift does not end the opportunity; the last wheel does.
A direction threshold is not a full-force threshold
The ±10% steering line decides which direction is stored. The recovered tire-force multiplier also depends on how far internal steering is turned.
In the measured high-speed ice samples the target is about 1 + s^1.5, where s is the internal steering amount without its sign: 0% → 1.0, 20% → 1.089, 60% → 1.465, 100% → 2.0. That is the high-speed case; graph 5 shows how less speed shrinks the boost. On the clean tarmac start of ANGULAR ↻ MOMENTUM, this icy multiplier stayed at its baseline 1.0. This multiplier acts on a tire-force term, not on vehicle speed. Steering near zero can temporarily reduce it even while the stored direction survives the 300 ms neutral timeout.
Slow cars get little or no boost
The boost above 1.0× is scaled by how fast the car moves sideways across the front wheels. That depends on speed and on the angle between where the front wheels point and where the car is going. Full steering is assumed here.
The target is 1 + curve(x) × s^1.5. x is the car's speed across the steered front wheels; the Stadium car's curve is 0 below 36 km/h of x and 1 from 108 km/h, linear between. At full steering the front wheels turn 45° from the car's nose, so a straight run and a 90° slide both put them 45° off the travel direction: no boost below about 51 km/h, full 2.0× from about 153 km/h. In a slide of about 45°, countersteering lines the wheels up with the travel direction (0°), so it gets almost nothing at any speed. With less steering, graph 4's boost is scaled by the same curve. The timer and stored direction do not depend on speed. How speed limits the recovered tire force →
Why setting the direction before a jump feels instant
Changing stored direction resets the tire-force multiplier to 1.0. The measured delayed landing stayed low for about 400 ms and reached 2.0 at about 600 ms. Airtime can use up this wait.
The blue curve is an illustrative interpolation through approximate landmarks from one measured full-steering landing (near 1.0 through 370 ms, about 1.05 at 390 ms, 2.0 near 590 ms). It is not the exact game recovery formula. The green line shifts that same sketch earlier by the selected lead time. Full steering and similar high-speed conditions are assumed. Neutral steering for about 300 ms is a separate stored-direction timeout, not this force-recovery time.
What the 300 ms neutral timer remembers
Start with left mode stored, then let internal steering stay between −10% and +10% while a wheel touches. Drag how long it stays there before returning left.
The graph shows stored direction, not the current force multiplier. Near-zero steering can lower the multiplier even before the stored mode clears. The 300 ms line is the timer threshold; the mode clears on the next eligible physics update at or after it. This simplified graph assumes continuous wheel contact and an active icy tire-force branch. In the air the neutral check does not run; at landing the timer may be checked if steering is still neutral.
Why the force dips after a plastic flick
One measured replay over a plastic flick, with full left steering and gas held throughout. The stored direction never changes, yet the multiplier falls from 2.0× to 1.0× for about 150 ms. As in graph 5, the target depends on the angle between the steered wheels and the direction of travel, and plastic changes the wheel angle.
Measured on 30 September 2026 on the tested build: every other Gorilla Grip Logger frame, plus the Gorilla Grip Trainer's logged force from the same replay. Both angles are measured from the car's nose on the same side. The wheel angle is 45° × steering × the icing curve from graph 1: the ice-family curve while any wheel touches Ice, Snow or RoadIce, and the plastic/other curve while all four are on plastic. On the plastic the tires lose icing and the plastic curve shrinks the wheel angle to about 18°, so the wheels stay far enough across the travel direction for 2.0× while the car swings round. When the front-left wheel reaches RoadIce at 8.28 s, the ice curve turns the wheels back to 43.6°, almost exactly where the car is going, and the target drops to 1.0×. It climbs back as the car keeps turning, with no timer involved. The material-selected prediction stays within 0.04 of the logged force; with the ice curve alone it misses by up to 0.9. The measurement in detail →