# Steering smoothness and corrections: steering rate and error metrics

> Steering traces show both driver technique and car balance. Steering-speed and steering-error metrics quantify smoothness and corrections – but only mean something once you know whether the car is balanced and the driver is at the limit.

Area: Driver Analysis · Starting level: L2 Practitioner · Updated 2026-11-16
Source page: https://race-engineer-academy.vercel.app/lesson/driver-steering-smoothness

## L1 · Foundation

**The steering wheel is a two-way channel.** The driver feels front grip through self-aligning torque in the rim and rear grip through yaw and lateral acceleration in the seat, and responds with steering inputs. A smooth trace means a car that does what it's asked; a ragged one means road feedback, balance problems, or an over-busy driver.

Basic shapes in a single corner:

- **Ideal:** one smooth build-up to a steering peak around the apex, a short hold, then a smooth unwind as throttle comes in.
- **Understeer:** steering keeps increasing while lateral g stays flat or falls.
- **Oversteer:** steering reduces sharply, sometimes beyond zero (opposite lock), often with small dips in lateral g.
- **Busy:** many small reversals of the steering rate, each a correction.

A driver below the limit needs fewer corrections, so a smooth, low-rate trace may simply mean the corner was taken slower than possible (Segers: steering-speed amplitudes fall when a corner is taken at lower speed). Before judging smoothness, confirm the driver is at the car's limit and the car's balance is acceptable – otherwise you are measuring the setup or the margin, not the driver. Compare steering rates corner by corner: slow corners need large wheel angles (the geometric term $57.3L/R$ grows as the radius shrinks) and naturally show higher rates.

## L2 · Practitioner

**Steering speed channel.** Two ways (Segers):

1. Differentiate: $\dot{\delta} = |d\delta_{sw}/dt|$ in deg/s.
2. Subtract a filtered version from the raw signal: $|\delta_{sw} - \tilde{\delta}_{sw}|$ – the high-frequency part only.

Take the absolute value so left and right turns count equally. The **lap average** is a smoothness statistic: lower = smoother. Segers' Interlagos comparison: the smoother driver had lower average steering speed and, in that dataset, faster laps.

**Corrections count.** Count reversals of steering rate with amplitude above a threshold (e.g., > 3–5° of wheel angle within 0.3 s) per corner or per lap. Map them on the track: corrections clustered at the same corner every lap are a car (or line) issue; scattered corrections suggest driver or traffic.

**Steering error metric.** Fit a smooth reference (low-pass filtered steering, or the steering predicted from lateral g and understeer gradient) and compute the RMS difference per corner:

$$e_{RMS} = \sqrt{\frac{1}{N}\sum_i\left(\delta_i - \delta_{ref,i}\right)^2}$$

Small values mean the driver tracks a consistent intention; large values mean constant correction.

**Typical numbers:** turn-in steering rates of 150–400 deg/s at the rim are normal for pro GT drivers in medium-speed corners; formula cars with quick racks use smaller wheel angles (often < 90° lock-to-apex) and correspondingly lower rates. Compare within one car only.

**Caveat (Segers):** smoother is not universally faster. On some tracks and conditions, a driver who rotates the car aggressively with steering is quicker. Use the metric to track a driver against himself and to spot changes after setup changes.

## L3 · Advanced

**Separating balance from technique.** Plot steering against lateral g for a corner as an XY. Only at constant radius does a steady-state, linear car give a line whose slope (road-wheel angle, i.e. steering-wheel angle ÷ steering ratio) is the understeer gradient; across corners of different radius the geometric term $57.3L/R$ dominates, so use the understeer angle instead (see the balance lesson). Hysteresis loops show transient behaviour: wider loops on entry = the car lags the driver's input (slow response, under-damped yaw); loops with reversal at the end = exit oversteer corrections. A setup change that reduces loop width usually makes drivers "smoother" without any coaching.

**Steering phase relative to lateral g.**
- Entry understeer: steering peaks *before* lateral g peaks and keeps the angle high while g lags.
- Mid/exit understeer: lateral g peaks *before* steering; steering keeps increasing while g decays.
- Oversteer: steering rate goes negative while lateral g dips.
Segers uses these patterns to read balance from driver activity (see the balance lesson).

**Noise sources.** Kerb strikes and road bumps kick the rack through bump steer and kickback; they appear as short high-frequency spikes in steering speed that the driver did not intend. Gate the smoothness metric away from known kerbs or use the filtered-difference approach with an appropriate window (e.g., 0.1–0.2 s). Power-steering systems (GT, TCR) filter feedback differently from unassisted formula cars, which also changes how drivers steer.

**Track-level metrics.** Average steering-speed per lap, total steering integral per lap (Segers shows lap steering integral rising with understeer and slower lap times across five cars), and corrections per lap are good run-chart metrics. Watch them across a stint: rising corrections often signal rear tyre degradation before lap time drops.

## Key formulas

**Steering speed**

$$\dot{\delta} = \left|\frac{d\delta_{sw}}{dt}\right|$$

- \delta_{sw}: steering wheel angle (deg)
- \dot{\delta}: steering speed (deg/s)

**High-pass steering activity**

$$\delta_{HF} = \left|\delta_{sw} - \tilde{\delta}_{sw}\right|$$

- \tilde{\delta}_{sw}: low-pass filtered steering angle

**Steering error (RMS)**

$$e_{RMS} = \sqrt{\frac{1}{N}\sum_{i=1}^{N}\left(\delta_i - \delta_{ref,i}\right)^2}$$


## Trackside takeaways

- Check car balance and whether the driver is at the limit before judging steering smoothness.
- Average |steering speed| per lap is a simple smoothness statistic – track it per driver, not across cars.
- Corrections repeating in one corner point to the car or the line; scattered ones to the driver.
- Steering vs lateral g XY loops reveal transient balance; setup can “make” drivers smoother.
- Smoother is often faster, but not always – use the metric as a diagnostic, not a goal.

## By car category

### GT3

Power steering and long wheelbase: steering traces are relatively clean; rising corrections late in a stint often signal rear tyre degradation.

### TCR

Front-drive cars show torque-steer and kickback on throttle; distinguish driver corrections from drivetrain effects at exit.

### FIA F3

Unassisted quick steering with small angles; corrections are short and sharp – sample steering at ≥ 100 Hz.

### One-make Cup

Identical cars make team-mate smoothness comparisons fair; use per-corner corrections maps in debriefs.

### IndyCar

On road courses without power steering in older specs, steering effort is high and fatigue can show as increased corrections late in a race; check the current spec for assist.

### LMP2

Endurance with Pro-Am line-ups: corrections per lap highlight which driver struggles with a balance – useful for compromise setups.

## References

- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §14.6 Steering
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.4 Driver Activities That Indicate Vehicle Balance
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §17.2 Why Use Metrics? (steering angle integral vs lap time)
