An asking price is an opinion. A sold price is a fact. That one sentence, applied without mercy, fixes most pricing mistakes in this hobby. Here's the full discipline behind it.
WHY ASKING PRICES LIE TO YOU
- Survivorship: the listings you see are, by definition, the ones that HAVEN'T sold. A watch priced right disappears in days; the overpriced ones accumulate and dominate your search results. Browsing active listings systematically shows you the prices buyers rejected.
- Anchoring: sellers ask high to leave negotiating room, and your brain anchors on the first big number it sees. Ten minutes of scrolling $600 asks makes a $450 buy feel like a steal — even when the reference sells for $380.
- Stale hope: plenty of asks are years old, orphaned listings from sellers who don't actually need to sell. They're not prices; they're monuments.
Solds are what someone actually paid, with money, recently. That's the only definition of price that survives contact with your own listing.
HOW TO BUILD A COMP SET THAT'S ACTUALLY COMPARABLE
Pulling "solds" isn't enough — pulling the RIGHT solds is the skill:
1. Exact reference. Not the model family — the reference. Variants (dial color, bezel insert, bracelet vs strap, regional versions) can carry meaningfully different prices.
2. Match the kit. Full kit (box, papers, links, tags) vs bare head is a different product. Never mix them in one comp set; you'll price your bare head against full-kit sales and wonder why nobody bites.
3. Match the condition tier. Unpolished-with-crisp-edges, honest-wear, and beat-up-project are three different comp sets. Be brutal about which tier YOUR watch is in — buyers will be.
4. Recency window. Prefer the last 60–90 days. A hot reference can reprice in a season; last year's solds describe a market that no longer exists.
5. Include what the buyer paid in total. On platforms where shipping is separate, a "$375 + $15 shipping" sale is a $390 comp. Compare totals to totals.
6. Note the sale format. A no-reserve auction ending at 6am Tuesday is real data but it's the bottom of the range; a Buy-It-Now that sat two weeks then sold is closer to a market-clearing price. Keep both, know which is which.
HOW MANY COMPS IS ENOUGH?
Rules of thumb, not laws:
- Five-plus recent, genuinely comparable solds: you have a price. Trust the cluster.
- Two to four: you have an estimate. Widen your margin of safety and lower your max buy accordingly.
- Zero or one: you have a story. Thin data is itself information — it means illiquidity, which means longer hold times and wider risk. Some of the best flips live here, but only priced for that reality.
MEDIAN, NOT AVERAGE — AND WHAT TO DO WITH OUTLIERS
Use the median of your comp set, not the mean. One $650 outlier in a $380–$420 cluster drags an average up $40 and tells you nothing; the median shrugs it off.
Outliers deserve a second look before you delete them, though. A sale far below the cluster usually had a flaw the photos showed and your summary didn't capture; a sale far above usually had something extra (rare dial, full documented service, original owner kit). Either way the outlier is telling you what moves price for this reference — that's negotiation ammunition.
A labeled illustration: last six solds for a reference are $355, $370, $380, $385, $395, $600. The $600 was a full-kit first-owner set; yours is a bare head. Your comp set is really the first five → median ≈ $380 → that's your expected sale, and everything (fees, shipping, margin) works backward from it.
WHERE ASKS ARE ACTUALLY USEFUL
- Direction: if every active ask sits far above recent solds, sellers believe the market is rising (or they're dreaming — the next few solds will tell you which).
- Competition check before YOU list: your listing will sit next to those asks. Price a hair under the plausible ones and you're the obvious choice on the search page.
- Ceiling sanity: your expected sale should almost never exceed the cheapest credible active ask — a buyer can just buy that one instead.
HOW THE DESK DOES THIS
The PWF deal engine is this exact doctrine in software: it wants verified sold prices in the comp library before it will bless a buy, it weighs the set and flags thin evidence, and the Chrome comp importer pulls sold-result cards straight off an eBay sold search so building the set takes seconds instead of a spreadsheet evening. And one thing you can't scrape from anywhere: when a member deal on our Deal Board completes, the sold price becomes a shared first-party comp for members.
Bring a reference you're struggling to price into this category and we'll work it as a group — that's what Pricing, Comps & Market Talk is for on the member side.