The receipts
Anyone can claim their estimates are accurate after the fact. Nestraq does it the other way round: estimates are locked into a tamper-evident ledger before homes sell, then settled against what they actually sold for — and we publish the whole distribution, misses included. This page is that scoreboard, plus exactly how it works.
Pre-committed estimates
Every estimate below is written into a tamper-evident hash chain before the home sells — hash-chained since 5 July 2026, with the chain head anchored publicly. Live receipts settle against HM Land Registry sold prices as sales complete (typically 3–6 months), and the settled results — misses included — are published here as they accrue.
Our estimates vs what actually sold
While the live chain accrues, this backtest shows how the same method performs: for homes sold in 2025, we rebuilt the estimate our method would have produced the day before each sale — using only sales completed strictly before that date (the cutoff is enforced in code) — then compared it with the actual sold price. Every miss is included; nothing is hand-picked.
Loading the scoreboard…
Contains HM Land Registry data © Crown copyright and database right. This data is licensed under the Open Government Licence v3.0. Estimates are indicative statistical figures from public sold-price data — not a survey, mortgage valuation or advice.
Why pre-commitment matters
A prediction only counts if it provably existed before the outcome. Each estimate snapshot is hashed together with the one before it (SHA-256), so rewriting any historical estimate breaks every hash after it. The chain has been running since 5 July 2026, and its head is anchored publicly — committed to a public repository on every refresh — so we can't quietly regenerate the record either. The chain head you see above commits to every estimate beneath it.
Snapshots are system-generated from our free scoring — a pseudonymous property key, the postcode district, property type and bedrooms, the estimate band and a timestamp. No exact addresses, and no user data of any kind.
How the backtest works
Live receipts take months to settle — a home flagged today sells next quarter. So while calendar time accrues, the backtest answers the same question retrospectively, under a strict rule: for each home sold in the backtest year, the estimate is rebuilt using only sales completed strictly before that home's sale date. The cutoff is enforced twice in code — a date filter, then a fail-closed assertion that aborts any estimate a post-sale comparable reaches. A backtest estimate can never have seen the sale it's marked against.
The estimator is the live method: same-postcode HM Land Registry comparables, same-type stratification, outlier trimming, a median point estimate with a 10th–90th percentile band, and the same confidence rule — only estimates the live product would actually publish (medium or high confidence) are settled. One deliberate difference: the backtest skips the time-indexing the live method applies, so if anything it overstatesthe live method's error. Backtest results are labelled BACKTEST everywhere and are never mixed into the live chain.
What we publish — and what we refuse to
- Distributions, never anecdotes. Median absolute error, the share within 5% and 10%, the worst miss, per-region breakdowns. No hand-picked success stories — the UK advertising regulator has upheld complaints against unrepresentative property case studies, and we think that standard is right.
- Misses included, structurally. The scoreboard is computed over every settled estimate; there is no code path that filters to the flattering ones. The worst miss gets its own tile.
- Nothing below the publication gate. Regions with too few settled estimates are omitted entirely rather than shown with false precision.
- Estimate-vs-sold only. Both sides of every settled receipt are self-generated or Crown open data. Listing history (asking prices, reductions, time on market) appears in aggregate only.
- No per-property distress signals — ever. Public artefacts never carry per-address insolvency or distress flags.
Check it yourself
The scoreboard on this page is rendered from a public endpoint you can query directly:
GET https://api.nestraq.com/receipts/summary
It returns the chain head, the LIVE pre-commitment count and the full BACKTEST distribution — the same numbers, no more, no less. The estimates behind them are the same free ones the app shows; the methodology page explains how they're built, the Overpricing Index shows what they add up to per area, and why no language model is ever allowed to set a number.
Daily radar calls are published as dated public receipts at /r.
Contains HM Land Registry data © Crown copyright and database right. This data is licensed under the Open Government Licence v3.0. Estimates are indicative statistical figures derived from public sold-price data — not a survey, a mortgage valuation, or financial or property advice.