Sold comps versus asking price: what the market actually paid

Asking price is a biased sample

The pool of active listings is survivorship-biased in the wrong direction: reasonably priced units sell and leave the pool, while overpriced units persist and accumulate. Averaging what remains systematically overstates value, and the bias grows with how long you look.

For slow-moving or collectible categories the distortion can exceed 30 percent. Any tool that quotes 'market price' from active listings is quoting the price of items the market has already declined.

The estimator

MarginMap computes fair market value from completed sales only, using a recency-weighted median rather than a mean.

Median, not mean

A single restoration-grade sale or a mistyped price cannot drag the central estimate.

Recency weighting

Sales decay exponentially over the window, so a shifting market shows up in weeks, not quarters.

IQR trimming

Values beyond 1.5 interquartile ranges are excluded and shown separately as flagged outliers.

Condition stratification

Comps are grouped by condition grade before aggregation; a used-fair sale never sets the price for a sealed unit.

Sample size is a first-class output

A median of four sales and a median of ninety sales are different objects and MarginMap refuses to render them identically. Below the minimum sample threshold the estimate carries a thin-data flag, the confidence interval widens visibly, and the buyer and reseller scores discount the data confidence factor accordingly.

Liquidity: price is half the answer

Two variants can share a median and behave completely differently. Sale velocity — how many completed sales the window contains and how tightly they cluster — determines whether that median is achievable in two weeks or two quarters. Days-to-sell is surfaced next to every median and feeds the reseller liquidity score directly.