Skip to main content
To say a driver was “fast” you need a yardstick: what did the field run on that lap? The field pace curve builds that yardstick. It plots the trimmed field-average lap time as it evolves over race distance. Each lap time is first corrected for fuel load and for the expected tyre compound cost, so what is left is a single reference baseline the curve every other decomposition measures against.

​
What it shows

The chart overlays a smoothed reference, the raw mean, and a caution flag:
  • Smoothed curve (field_pace_smoothed_s): a 5-lap centred rolling average of the trimmed mean. The window never crosses a safety car, virtual safety car or red flag lap, and the first two green laps after a safety car or red flag keep their own trimmed mean. This is the primary answer to “what did the field actually run this lap?”
  • Raw trimmed mean (field_pace_trimmed_mean_s): the mean corrected lap time per lap number after dropping the fastest 10% and the slowest 10% of eligible laps (percent rank 0.10 to 0.90). The ribbon between the two series shows where smoothing diverges from the raw data a wide ribbon means a noisy lap (few cars, spread strategy).
  • Low-sample laps (low_sample_flag): lap numbers with fewer than 5 eligible laps (pit windows, early laps) are flagged. A flagged lap has no smoothed value.
The smoothed curve is the base_track_pace_s input to the additive decomposition so this is not just a chart, it is an upstream model input.

​
How to use it

  1. Select a race from the filter bar; the curve loads.
  2. Read the smoothed curve as the field reference for that race.
  3. Watch the ribbon width where the smoothed and raw series diverge, the lap is noisy and the average is less reliable.
  4. Discount flagged low-sample laps pit windows and opening laps have too few cars to average well.

​
Reading the signal correctly

The curve is a trimmed central tendency, not the leader’s pace or the fastest lap. Per lap number, the model drops the fastest 10% and the slowest 10% of eligible laps before averaging. Fuel load is removed from every lap time before averaging. The expected tyre compound cost from the tyre cliff model is also removed before averaging. So a downward slope in the curve is not fuel burn. The model source (int_field_pace_curve) names the remaining trend the rubber gradient: track rubbering-in plus anything the fuel and compound corrections do not capture. The curve does not split that remaining trend into separate causes.
Early laps and pit windows are sparse and flagged for a reason the curve there leans on a handful of cars and can swing. Don’t read fine structure in the flagged regions.

​
Data source

The chart queries int_field_pace_curve via DuckDB-Wasm in the browser, for a single race. This is a single-race view across 2018–2025 data. For the full model definition, see the int_field_pace_curve reference.