We started with 841 collector models that clear a price floor and carry enough sold results to price honestly. 518 of them were gated out: too few sales, a bucket so wide that one fair price would be fiction, or a nameplate our own records cannot separate cleanly. That left 323 cars, and the rules made 25 calls out of those. The desk added 12 Buys of its own, each marked as a desk call with our model's view printed beside it.
Each one is scored on the same seven measures: what the forecast says, how confident it is, how many cars are for sale against how many have sold, where recent hammer prices sit against our own estimate, what the same chassis returned when it sold twice, the twelve month move, and the gap between dealer asking and the last print. Then the desk writes the reasons, and where it disagrees with the score it says so in the report.
A Buy needs high forecast confidence. A Keep or an Avoid may run on medium. Every Avoid speaks about a model, a specification and a price gap. Never about a seller, and never about a specific car for sale.
The list leans Ferrari. That is not taste, it is evidence: Ferrari has by far the deepest sales record of any marque we track, so Ferraris are the cars with enough history to clear the confidence bar. The report says so on its method page too.
What it is built on
- Sold results on file
- 78,374
- Sold in the last 12 months
- 22,716
- Models priced
- 1,951
- Models with a live forecast
- 1,546
- Forecast error in backtest
- 18.8%
- Chassis watched trading twice
- 1,247
Forecast error is the median absolute error when the model is tested against sales it never saw. Figures as of 2026-09-24, prices in USD.