How every list is built — sources, formula, guarantee.

1. Sources (all free, all official)

2. Scoring formula (published, frozen at build time)

depth_years   = sale_year - first_delinquent_tax_year      (min 1)
equity_ratio  = (assessed_value - total_owed) / assessed_value
score         = 10 * depth_years * (equity_ratio > 0 ? equity_ratio : 0)
                + 90 * (equity_ratio >= 0.9 ? 1 : 0)        # deep-equity bonus
flags         = owner_is_entity, tax_sale_indicator, partially_paid (published columns)

Scores are computed once per county roll and never edited afterward — the same freeze rule used on the public prediction ledger.

3. Comping engine backtest

The ARV model behind our value columns was backtested on 57,985 real recorded sales (New York City (geography-agnostic model; CO/VA re-tune in progress)): 7.1% median absolute error, 71.5% of estimates inside the ±10% band. Zone A re-tune on Colorado sold-deed data is in progress and the numbers on this page update when it lands.

4. The guarantee (replaces refunds)

• Every row ships with its source parcel ID and the name of the source system.
• If >10% of rows fail verification against the official record, you choose: replacement file or credit back.
• Data is point-in-time: counties correct rolls after publication. Rows you can't reproduce against today's roll aren't our miss — the build date is printed on every file.

5. What we never do

• No cold outreach to anyone on these lists — we sell data, we don'solicit the owners.
• No credit-header/FCRA-shaped data, ever.
• No scraping behind login walls; if a portal blocks automated access, we use the alternative official source and say so.

← Back to the Denver list