SGCarMart Listings

Browse the structured CSV as a website table with search and pagination. Search checks the whole row, so titles, descriptions, brands, fuel types, and dates are all fair game.

Loading dataset...
Valuation Methodology

This viewer uses a comparable-listings method, not a machine learning model. Each car is compared only against listings in the same market bucket.

  • Each listing gets a comparable_key based on normalized title, COE category, and powertrain group.
  • Strict comparables must share that key and be reasonably close on registration year, COE left, mileage, engine size, owner count, and fuel type when available.
  • The checker tries strict matches first, then a slightly relaxed pass in the same bucket only. There is no brand-level fallback.
  • A listing needs at least 5 usable comparable prices before it can be scored.
  • The fair price is the trimmed median comparable asking price, stored as estimated_price_sgd.
  • price_gap_pct is the difference between the listing price and that estimate.
  • The label is undervalued at 15% or more below estimate, overvalued at 15% or more above, and otherwise fairly-priced.
  • If the comparable set is too weak or too noisy, the result is downgraded to insufficient-data.

This is a relative asking-price signal, not a true market-value appraisal. It does not know actual transacted prices, condition, accident history, warranty, dealer quality, or modifications.