Everything we publish about how California tax-deed sales actually work — the failure modes, the legal mechanics, and an honest account of what our pipeline does and doesn't do. We write these to demonstrate the thinking behind the sale briefs, not to sell a course. Where we cite the law, we cite the primary source. Where we got something wrong or unfinished, we say so.
Worked examples · The traps · How the pipeline works
The cornerstone posts — read these first.
How the pipeline works
Anyone selling you a neural network to find paper lots is selling you a neural network to read a field that says NONE. The honest division of labor: rules cut the list, AI does the analyst work, you decide.
Worked examples
946 parcels in, 175 out. The real per-filter counts from a closed Riverside County tax sale — 687 paper lots gone on a single field check — plus an honest account of the valuation step, which has since been finished across all 175.
The traps
Every trap leaves a tell in the raw county list: a missing street address, one owner on thirty rows, a golf-themed street name, a non-baseline TRA code, an 'auto' in the LLC name. What each means, with the code sections.
The traps
Between a county approving a tax sale and holding it, owners redeem and parcels vanish from the list. We measured which ones, in two counties. On price, redemption is indistinguishable from random — the parcels that leave have the same median bid, the same assessed value, the same ratio as the ones that stay. On one thing it is not random at all: whether something is built on the land.
Worked examples
Kern County sells 1,017 tax-defaulted parcels on September 14–16. We read the sale months early out of the county's own approving resolution, then checked it against the county's live list when that landed: 203 parcels redeemed in between. On price the redeemed were indistinguishable from the survivors — but 22% of them carried a structure, against 7.5% of what is left. The median parcel opens at 1.34× assessed value.
Worked examples
Los Angeles publishes something almost no other county does — the year each parcel first went tax-defaulted. Join it to the assessor's values across all 1,252 parcels in the October sale and the pattern is almost mechanical. Among parcels the county assesses above $10,000, fewer than one in a hundred of those defaulted under ten years opens above assessed value. Of those defaulted thirty years or more, 98.7% do. The property is not what moved.
The traps
When a parcel doesn't sell at a tax auction the county offers it again, often cheaper. That reads like a bargain and usually isn't. Inside a single Fresno sale, the parcels being re-offered open at a median of 5.96× the county's own assessed value — the ones offered for the first time open at 0.13×. Placer cut its Truckee minimums 42% and they still open at up to 49× the county's figure.
Worked examples
Mariposa County noticed 24 items for its August 6, 2026 auction; ten were redeemed and one pulled, leaving 13. We matched all 13 to the county assessor's own parcel records. Every one opens below its assessed value, between 0.11× and 0.58× — the reverse of San Bernardino, where most parcels opened above it. All 13 also carry special assessments the notice calls a lien. A free pre-sale desk analysis, published while the parcels are still live.
The traps
We published a finding — 72% of the priced parcels in San Bernardino's July sale open at or above the county's own assessed value — and got told, fairly, that Proposition 13 explains it. It doesn't. So we went back to our own numbers and regressed opening bid against assessed value. The bid barely tracks value at all: within one tax-rate area it's a large fixed amount plus about 16 cents per dollar of assessed value. Here's what that means for reading any tax sale, and what it means for our own headline.
The traps
A tax sale is supposed to collect a debt, not pocket a windfall. Until 2023, in a lot of states, it did both — a county could sell a home over a small tax bill and keep every dollar above it. Then the Supreme Court, unanimously, said no. Here's what Tyler v. Hennepin changed, why it matters if you buy at these sales, and where California already stood.
Worked examples
San Bernardino County is auctioning 1,614 tax-defaulted parcels July 11–17, 2026 — mostly raw desert and mountain land, plus 364 timeshares. For about 72% of priced parcels the opening bid meets or exceeds the county's own assessed value, because the cost of putting a parcel through a tax sale is close to fixed and most of this dirt is assessed below it — the parcels where the ratio means something are a minority of 148. A free, pre-sale desk analysis, with the method shown in full.
Worked examples
Santa Cruz County listed 27 tax-defaulted parcels on Bid4Assets, June 5–8, 2026: 14 sold, 4 drew zero bids, 9 were withdrawn before close. Every winning bid, captured after the fact — from a $115,556 close on Glenwood Drive to four 1911 'Happyland' lots nobody would take at $1,000.
How the pipeline works
Anyone selling you a neural network to find paper lots is selling you a neural network to read a field that says NONE. The honest division of labor: rules cut the list, AI does the analyst work, you decide.
Worked examples
946 parcels in, 175 out. The real per-filter counts from a closed Riverside County tax sale — 687 paper lots gone on a single field check — plus an honest account of the valuation step, which has since been finished across all 175.
The traps
Every trap leaves a tell in the raw county list: a missing street address, one owner on thirty rows, a golf-themed street name, a non-baseline TRA code, an 'auto' in the LLC name. What each means, with the code sections.
No research matches that filter yet.