Methodology

How every number on this site is produced, and how accurate it is.

What “fair market value” means

The price a typical private-party sale of the watch would reach today: pre-owned, with box and papers, at the market-wide average across regions. Everything else on a watch page (new or unworn, without box and papers, a region, the dealer price, the typical asking price) is derived from that value by effects the model learns from the data.

Accuracy

Accuracy is published once enough sales have been held out to measure it.

How the value is built (fmv-v1)

1. Every price is evidence, weighted by what it is. Completed sales, auction results, listings marked sold without a price, and asking prices all count, but with different reliability. The model learns how noisy each kind is and weights it accordingly, so one sale outweighs several asks.
2. Adjusting for what differs (hedonic adjustment). A regression compares copies of the same reference traded in the same month, so it isolates what box and papers, condition, region, dealer vs private sellers, and the listing’s age are each worth. Asking prices are adjusted down to sale value, and listings that have sat unsold longer are treated as more over-priced.
3. Following the value over time (state-space model). Each watch’s value is tracked by a Kalman filter, a statistical way to update a running estimate as new evidence arrives. New prices move it in proportion to how reliable they are and how uncertain the value already was; with no new prices, the value holds while its uncertainty grows. Each day’s value uses only data available that day, so history is never rewritten with hindsight.
4. Guarding against bad data. Prices more than four times away from a watch’s typical price are dropped as data errors, and prices that sit far outside what the model expects are set aside once a value is established. The same watch cross-posted on several sites counts once.
5. Publishing only when confident. A value appears once it rests on at least three prices and its uncertainty is below about 15%. The shaded range on charts is where 80% of private sales are expected to land.

Watch metrics

• Price change (1M, 3M, 6M, 1Y): today’s fair market value against the value on that earlier date.
• Value vs retail: fair market value divided by the latest retail price.
• Active listings: listings seen in the last 30 days. Sales (1Y): sold listings and auction results in the last 365 days.
• Days on market: median time from first seen to sold, over the last year’s sales.

Indexes (chained-tv-weighted-v1)

• Baskets: the Overall index uses all watches; brand indexes use one brand (published once a brand has 5 priced watches); price-band indexes use watches whose value falls in the band on the rebalance day.
• Members and weights: on each rebalance (January 1, and the index’s first day), the top 300 eligible watches by trailing-year trading value. Trading value is the total of sales in the last 365 days; for watches with no sales yet we use value × active listings.
• Daily value: the index moves by the weighted average of members’ daily value changes. Base: 1,000 on the first day with at least 5 priced members.

Data

Prices come only from sources we own or are licensed to use. Listings are matched to watches by reference number. Non-USD prices are converted with European Central Bank reference rates for the day the price was seen (Source: European Central Bank).