What's a residual?
Residual = actual price − expected price.
"Expected price" comes from a least-squares regression fitted on the currently filtered subset:
expected_price = intercept + b₁ × mileage + b₂ × age_months
- Negative residual → priced below what comparable miles/age cars sell for (potential bargain).
- Positive residual → priced above what miles/age would suggest (overpriced).
- Zero → priced exactly on the regression line.
Worked example. A 12-month-old, 10,000-mile Octavia listed at £20,000. Suppose the regression on the filtered subset says a car at those miles & age "should" sell for £22,500. Residual = 20,000 − 22,500 = −£2,500 — i.e. £2,500 cheaper than its peers.
Caveats. The regression uses only mileage and age, so a high-spec vRS at 10k miles will look "cheap" compared to a base SE at 10k miles even though the comparison is misleading. Filter to a single trim & body type before reading residuals. The metric also ignores condition, options, colour, and location bias — things dealers price on but the listing data doesn't expose. Residuals recompute every time you change a filter, so the "expected" benchmark always reflects what's currently visible.