About GrailWatch
GrailWatch is an independent reference for high-end watch reliability, quality, and value, with under-market discovery from trustworthy sellers as a secondary layer. We're independent, we have no inventory, and we're not a dealer.
The clearest one-URL answer to "what does GrailWatch actually do?" is the worked example of a hypothetical Speedmaster Professional 3861 listing — landed cost, band, deal tier, trust read, risk read, verdict. Every step mapped to the open-source engine code.
What we do
We collect public marketplace and auction-archive data, run it through a deterministic scoring pipeline, and publish what we found. Every listing we surface gets:
- A fair-value band built from comparable sales, adjusted for condition, box & papers, brand tier, and era.
- A landed-cost estimate (sticker plus duties, brokerage, VAT, third-party verification fees) — because the right price comparison is total cost, not the ad price.
- A three-layer trust read (platform, listing, seller), each scored independently and combined transparently.
- A graded risk read (clear, caution, elevated, high) that names factual signals — never accuses anyone of fraud.
What we won't do
- Call anyone a fraud or a scammer. Risk is graded, based on factual signals. The word "fraud" is code-enforced out of published copy.
- Take payment from sellers to rank them higher. The merit ranking and any monetization (affiliate, membership, ads where present) are architecturally separate.
- Republish copyrighted dealer content. We link to sources with attribution; we don't scrape paywalled material.
- Make claims we can't defend. Every published score carries provenance plus an audit trail of inputs.
How the score works
The valuation engine builds a 20th/50th/80th-percentile band from comparable verified sales within a 12-month window, weighted by authenticity confidence. The listing's stated condition and box & papers scale that band; brand tier and era pick the right factor table. A score below the band's lower edge is a "great deal"; above the upper edge is "overpriced." Bands are widened when the comp sample is thin; confidence is reported alongside the score.
Trust is the median of three independent layers — platform integrity, listing integrity, seller history — each on a 0-100 scale, with the composite never overruling a fundamentally failing layer. Risk is a score derived from named, factual signals (no comp coverage, age of seller account, stock-photo detection, off-platform payment requests, etc.).
Where the data comes from
The current public demo runs against curated seed data so the engine surface can be evaluated end-to-end. Production ingest from the eBay Browse API and selected auction archives is the next slice; details are in the ingest-worker spec.
Try it yourself
Paste the basics of a listing into the Deal Calculator. No signup, no wall — you get a band, a deal tier, a landed-cost estimate, and the risk read in a single pass.
Who's behind it
GrailWatch is built and operated by Brett Morrison. The product is independent and self-funded; if and when we run affiliate partnerships or display ads, every monetization channel will be disclosed in the legal pages and reflected in the live code, not in a buried footnote.
Reach us
- Security disclosure: see /security
- Privacy questions: privacy policy
- Acceptable use & content reporting: AUP