How we calculate Rolex prices
Every range on a reference page comes from recorded sold results, adjusted to one standard watch and weighted toward recent sales. This page explains each step in plain words. Method version fv-2.0, current price data from 2026-10-26.
What the price on a reference page means
The headline range describes one configuration: pre-owned, box and papers, private sale. That is a worn but complete watch with its box and papers, sold between private parties at the level of recent sold results. It is an estimate from market data, not an appraisal and not an offer to buy.
Today 460 of the 793 Rolex references we track have a published range, and 208 of those are at high confidence. The rest show "not enough data" and list the individual sales we have.
Which sales count
- Only sold results: eBay sold listings and auction results from 18 vetted sources, from the last 24 months.
- Asking prices never count. A price someone hopes to get is not a price someone paid.
- Forum posts marked sold are shown separately on reference pages, labelled as the listed price when marked sold. They do not feed any number.
- Each sale keeps its date, condition, box and papers and source, and our archive holds 108,034 sold results in total: 87,061 eBay sold listings and 20,973 auction results. Forum posts are left out of that figure.
- Obvious errors and special pieces (unusual dials, gem settings, wrong references) are excluded by written rules before any calculation.
Adjusting each sale to the same watch
A watch sold without its box and papers, or one that was never worn, does not sell at the same level as a worn full set. Instead of guessing a percentage, we measure the difference in our own sales: each sale is compared with sales of the same reference within 90 days (at least 5 of them), and the typical gap is taken per collection, pulled toward the site-wide figure when a collection has few sales (a weight of 30 sales).
- On eBay, a watch sold alone went for about 13% less than a full set (531 watch-only sales against 15967 full sets).
- On eBay, a listing that does not say what comes with the watch sold for about 4% less.
- At auction, a watch sold alone went for about 8% less than a full set.
- An unworn full set sold for about 5% more than a worn one (2132 unworn sales).
We also measured the gap between auction and eBay prices. Adjusting for it did not make predictions better, so it is not applied: auction sales keep their own level. The "watch only" figure on a reference page uses the measured watch-only gap. It is shown only at moderate or high confidence and only when at least 100 watch-only sales stand behind the gap.
Recent sales weigh more
A sale's weight halves every so many days (its half-life). Busy references move faster, so their half-life is shorter. The half-lives were chosen by testing each option on a past year of sales and keeping the one with the smallest error.
| Sales in the last year | Half-life |
|---|---|
| 100 or more | 30 days |
| 20 to 99 | 180 days |
| Fewer than 20 | 180 days |
From sales to a range
The middle of the range is the weighted median of the adjusted sales: half of the weight sits above it and half below. The range around it is the spread of the middle half of those sales, never narrower than a minimum width that depends on the confidence level: plus or minus 25% at high confidence, 35% at moderate and 35% at low confidence (the overall floor is 15%). Those widths are the narrowest at which past ranges kept their promise.
What each confidence level promises
A range gets a confidence level from five checks. High confidence needs every check in the first column, moderate every check in the second.
| Check | High | Moderate |
|---|---|---|
| Effective number of sales after the time weighting | at least 8 | at least 3 |
| Age of the newest sale | at most 60 days | at most 240 days |
| Sales in the last 90 days | at least 3 | any |
| Spread of recent prices | at most 20% | at most 40% |
| The reference's own past prediction error | at most 15% | at most 30% |
We test the promise the honest way: each sale in the past year is predicted only from sales before it, and we count how often the later sale landed inside the range (test period 2025-10-27 to 2026-10-26).
| Confidence | Later sales inside the range | Predictions tested |
|---|---|---|
| High | 91% | 25632 |
| Moderate | 83% | 8673 |
| Low | 72% | 464 |
When we publish no range
A reference needs at least 5 sold results in the last 24 months, 2 of them in the last 12 months, from at least 3 separate sales, with the middle half of prices spread no more than 60% of the median, and an effective number of sales of at least 3 after the time weighting. Below that, the page says there is not enough data and lists the sales we have. We never fill the gap with a guess.
The volatility label
Next to a price you may see Stable, Moving or Volatile. It measures how spread out the last 90 days of adjusted sales are: Stable below 10%, Moving between 10% and 20%, Volatile above 20%. With fewer than 6 sales in that window there is no label.
Collection price indexes
Our collection price-history pages, and the market index in our blog reports, show prices as index points rather than amounts, with the first year set to 100. They are built in three steps, so a change in the mix of references sold cannot move the line on its own:
- Each reference contributes one series, its summary series or else its largest, so a comparison never mixes auction, forum and eBay prices.
- A reference's level for a year is the median of that year's monthly or quarterly medians, and it needs at least 6 sales that year.
- A year's move is the median change across the references with a level in both that year and the one before, at least 3 of them, and the moves are chained from 100. The current year is marked "to date".
Indexes are historical context only: no index feeds a price on this site. Forum figures in them are listed prices when a post was marked sold, not confirmed sale prices.
Updates and corrections
Prices are recomputed from the archive whenever new sales are added, and every page is rebuilt from that run. When the method changes, its version changes and this page changes with it. When we find an error in a sale record or a page, we correct the record and the next run corrects every number that used it. Who we are and how we stay independent: About WristPrices. To check a figure, start from all Rolex references.