How we make our verdicts
The comparable-home rule, the win model and its backtest, savings and tax estimates, and their limits.
Every number on this site comes from appraisal district records, computed by rules written down here and applied the same way to every home. Where a rule involves judgment, we say so. We describe what happened to similar homes last year; nothing here is a promise about your home.
1. Similar homes
Texas lets you protest a value above the median of comparable homes (Tax Code §41.43(b)(3)). We pick them by a fixed rule that never looks at values:
- Eligible: the same appraisal neighborhood and quality class, living area within ±15% and built within ±5 years of your home.
- Most similar first: eligible homes are ranked by physical similarity (a 5% size difference counts about as much as 2 years of age or a pool mismatch; the same subdivision breaks ties) and the15 most similar are kept.
- Too few: with fewer than 5 eligible homes, the verdict is "We can't tell yet".
2. Adjustments and the comparison
We compare house (improvement) values, not total values, because land is valued separately. Each similar home's house value is adjusted toward yours for size (value scales with the size ratio to the power0.66), a pool (8.4%) and age (0.5% per year), coefficients estimated from Collin County's 2026 values. The gap is how far your house value is above (or below) the median of the adjusted values. When at least 3 neighbors share your exact floor plan (same class and living area), we also show their median.
3. The verdict
Two things predicted a successful protest in Collin County's 2022–2026 results: the gap above similar homes, and how much the value rose from the year before (up to about 10%; bigger jumps did no better). A logistic model turns them into two chances: of any reduction, shown as "X out of 10", and of a meaningful reduction (about $100 a year or more in tax), which sets the verdict:
| Yes, worth protesting | chance of a meaningful reduction at least 65% |
|---|---|
| Maybe | 50% to 65% |
| Probably not | below 50% |
| We can't tell yet | fewer than 5 similar homes |
Honest limits. The model is fitted on about 84,000 Collin County protests filed by owners without an agent, 2022 to 2026. We tested it the hard way: fitted on 2022–2025 only, then scored on 2026 outcomes in neighborhoods it had never seen. It ranked homes only modestly (an AUC of 0.60 for any reduction and 0.58 for a meaningful one, where 0.5 is a coin flip and 1 is perfect). Homes it rated higher did win more often: in that test, owners whose homes it rated "Yes" won a meaningful reduction 72% of the time, "Maybe" 66% and "Probably not" 54%. But it can't see your home's condition, recent sales or how a hearing goes. Win rates also move from year to year (53% of these owners won a reduction in 2022, 74% in 2026), and the chances describe last year's results, so a new year can come in higher or lower. That's why we show frequencies ("7 out of 10") rather than precise percentages.
A home whose value looks high still gets "Probably not" when the tax math shows a lower value wouldn't lower this year's bill: a fully exempt homestead, or one whose 10% cap already holds its taxable value further below market than a protest could reach.
4. Protest outcomes
Collin County's export says, for each property, whether it was protested and how the protest ended. We checked its codes against the county's separate protest-case records: C marks protests closed without a hearing (mostly value agreements, plus withdrawals and dismissals), H marks protests decided by the appraisal review board, and A marks protests still open.
- Protested: any of the three.
- Decided: C or H on a certified value. Withdrawn and dismissed protests count as decided without a reduction; open ones are left out.
- Got a reduction (won): decided, with the certified market value below the noticed market value.
- Cut: noticed minus certified market value, also as a share of the noticed value.
- With an agent: a tax agent is on file for the property (in Dallas, or filed this year's protest).
Dallas County publishes protest results in separate appraisal review board files: each protested home's noticed value, the value it ended at, and whether the owner or an agent filed. We apply the same definitions: a protest settled, withdrawn or dismissed counts like C, one decided by a review board panel like H, and an open one like A. A home with no protest was noticed at its certified value. Dallas verdicts use the model fitted on Collin County's results until we fit one on Dallas's own.
Denton and Tarrant counties' exports don't include noticed values or protest results, so their verdicts use the comparison with similar homes only, with no "X out of 10" or savings, and pages say so. Where a page covers a county without outcomes, outcome findings use all the counties that publish them and say "Texas".
5. Savings and tax estimates
- Typical savings (with "X out of 10"): the middle half (25th to 75th percentile) of the cuts that winners in your neighborhood got, as a share of their value, applied to your home and priced in your yearly tax. Shown only for Yes and Maybe, and only with at least 5 winners nearby.
- Value reduction: the gap to identical floor plans when there are enough of them, otherwise the gap to the similar homes' median; the range adds what neighbors won.
- Tax: computed per taxing unit from the rates and exemption amounts the district publishes with its data, your exemptions (homestead, over-65, disability, disabled veteran), the 10% cap, and over-65 or disability tax ceilings, whose units are left out of savings. Where a district publishes no rates with its data (Dallas, Denton, Tarrant), we show no tax estimate.
Every estimate is a point and a range. Estimates are not promises: each protest is decided on its own evidence.
6. Area pages
County, city and neighborhood pages add up the same per-home results. Neighborhood names come from the most common subdivision name in the district's records. Streets are listed only with at least 10 homes; we never single out a house or show an owner's name.
See where our data comes from for sources, refresh schedule and the JSON API.