Sign convention is consistent everywhere in Off The Pace: negative is faster (shown in green), positive is slower (shown in red).
Pace delta
pace_delta is how fast a driver is relative to the field on a given lap not the raw lap time. A driver running at exactly the field’s typical pace has pace_delta ≈ 0. This normalisation is what lets you compare a residual at Monaco directly against one at Monza, even though the two circuits’ lap times differ by 20+ seconds.
The seven components
Every lap’space_delta is split into seven named pieces that sum to it exactly: fuel, compound (tyre wear), rubber (track evolution), ambient (temperature), constructor (the car), dirty air (aero wake), and driver skill (everything else). See the Transform tab for the full equations and what each term captures.
The tyre cliff
The lap at which a tyre compound’s grip falls off sharply tyre wear before the cliff is gradual; after it, pace loss steepens fast. The app’s Tyre Cliff Survival feature and Degradation Simulator both surface this onset point, fitted historically and predicted live by the ML models.Clean lap
A lap eligible for fitting the physics terms not the first lap, not a pit-in/out lap, not under Safety Car or VSC, not wet, not flagged as a data-quality anomaly. Dirty laps would bias every coefficient downstream, so they’re excluded before any term is estimated.Structural pace / ghost car
A constructor’s or driver’s pace with the confounds removed e.g. a driver’s speed independent of the car, or a car’s speed independent of the driver. The app’s Ghost Race Standings feature re-runs every result with this structural pace in equal machinery, answering “who would have won in the same car?”Stint and compound
A stint is the set of consecutive laps a driver runs on one set of tyres between pit stops. A compound is the tyre type (Soft, Medium, Hard, Intermediate, Wet) each with its own degradation curve and cliff onset, stored per circuit and season.Where the depth lives
Transform
Every equation, every coefficient, the CI-enforced identity.
Glossary
Precise, column-level definitions for every term.
Machine Learning
How the cliff and degradation predictions are trained and validated.