How this works

1What goes in

The site starts from three kinds of information:

• Official Dallas County early voting reports: the county's live vote-center counts, its daily turnout reports, and its daily lists of in-person early voters and returned mail ballots.
• Public voter registration records, matched with a private, proprietary voter database.
• Past election results for Dallas County, mainly the 2022 and 2024 November elections.

Those inputs feed a proprietary statistical model owned by Cinder and Crown LLC. We publish its outputs, when it ran, and how it has done. We don't publish how it works inside, the data it uses, or where that data was obtained.

2Vote counts

“In-person early votes” is the county's own running count of voters checked in at early-voting centers, shown as the county posts it. “Total early votes” adds mail ballots the county reports as returned. These are counts, not estimates, but they are still the county's unofficial figures and can be revised.

3Mail ballots

• Mail figures are ballots returned to the county, taken from the county's daily mail-ballot returns report.
• A returned ballot is not yet counted or tallied. Mail ballots are verified and counted later under the county's normal process, and some returned ballots may not be accepted.
• The figures update when the county posts a new report, usually once a day, so they can lag the live in-person count.
• The comparison with 2024 lines up the same number of days before Election Day. It's shown only when we have comparable 2024 figures; otherwise it's left off rather than guessed.

4Turnout comparisons

“vs. 2024 same day” compares this year's early vote with the 2024 presidential election (and 2022 midterm) at the same day of early voting. 120% means 20% more votes than at that point in 2024. Days are matched by early-voting day number, and the past figures come from the county's final published reports for those elections.

5The Early Vote Tide

• The Tide is our estimate of who has cast the ballots so far (in-person plus mail), grouped by voters' usual lean: R-leaning, Swing (voters without a clear lean, including voters we could not place) and D-leaning. It is shown in whole percentage points.
• Below the bar, the modeled split estimates how those same ballots would divide between the two parties once swing voters are allocated. The shift and the trend chart are based on the modeled split.
• It is not a count of anyone's vote. Ballots are secret and are not counted until the polls close; nobody, including us, knows how any person voted.
• The “shift” compares this year with the same point of early voting in 2024, in the same areas. “R +2” means the estimated mix is 2 points more Republican-leaning than at that point in 2024; “D +2” means 2 points more Democratic-leaning.
• Districts are shown only when enough ballots have been cast there that no individual could be singled out.
• Early voters are not a sample of all voters. Who votes early, by mail or on Election Day differs, so the Tide is a signal about early voters, not a prediction of the final result.

6Forecasts and win chances

• Each race shows the model's middle estimate of the Republican candidate's share of the two-party vote, with ranges. The 80% range is where the model expects the result to land about 8 times in 10. Results can fall outside it.
• Statewide and multi-county races show the Dallas County portion only. For U.S. Senate, Governor and other statewide offices, and for districts that cross the county line (some congressional, state senate, State Board of Education and appeals court districts), every share, range and chance on this site describes the votes cast in Dallas County only, not the overall outcome. Those races are tagged “Dallas Co. only”, and their chance is labeled as the chance the Republican carries Dallas County, not the chance of winning the race. A candidate can lose Dallas County and still win statewide, or the reverse.
• “Win chance” is the model's estimated chance that the Republican candidate finishes above 50%. A 70% chance still means the other candidate wins about 3 times in 10. Very small or very large chances are shown as “under 1%” or “over 99%”, never as certainties.
• The model combines what early voting shows so far with past results and a consensus of published public forecasts. It changes as new county data arrives.
• Projected final turnout is a modeled range for the total number of ballots, with the split by vote type.

7aDaily results

After each early-voting day, we save that day's official county counts (in-person plus mail ballots returned), the in-person comparison with the same day in 2024, the Early Vote Tide, and the forecast exactly as it was published. The arrows show how each race's modeled Republican share and win chance moved from the day before. Past days are not revised afterward, except to fix confirmed errors.

7bRace insight

• Turnout so far compares ballots cast so far in the race's district with the same point of early voting in 2024. For a district, we start from the county's official comparison and adjust it by how the district is running relative to the whole county.
• Modeled lean is the Early Vote Tide for that district, in whole percentage points.
• Win chance and trend come from the forecast; the trend is the change in the modeled Republican share since the first day of early voting.
• District figures appear only for the county and for districts with at least a thousand ballots cast. Countywide offices and courts use the countywide figures.

7Turnout map and vote centers

The turnout map shades each early-voting center by its turnout so far compared with the same day in 2024. It shows turnout only. No estimate of how a center's voters lean is shown for any individual center.

8Count Watch and transparency checks

Count Watch compares the county's public reports with each other, for example whether two reports of the same count agree and whether files posted on time. A status other than “Matches” means two public numbers haven't lined up yet, usually because of reporting schedules. These checks are not allegations or findings about any voter, election worker, official or office. See the Terms of Use.

9When numbers update

• Live in-person counts: from the county's public feed, refreshed at most every 2 minutes.
• Mail returns, the Tide, forecasts and the daily archive: updated once each morning (around 4 AM Central) after the county posts the previous day's reports, and again if the county posts corrections.
• Each model output shows when it was last run.

10No individual voter information

We never publish any individual voter's name, address, ID number, voting history or any estimate about a person. Published figures are totals for the county, large districts or vote centers, and estimates by lean are shown only for the county and for districts with large numbers of voters. Voter records used for the model stay on secured systems and are not shared. See the Privacy Policy.

11Limits and disclaimers

• Everything here is unofficial and for information only. Estimates are provided “as is”, without warranty of any kind, and can be wrong.
• The site is independent. It is not affiliated with or endorsed by Dallas County Elections, Dallas County, the Texas Secretary of State or any government agency.
• Nothing on this site is legal advice. For where, when and how to vote, and for official results, use dallascountyvotes.org and VoteTexas.gov.
• Full terms: Terms of Use.

12Corrections

If you think a number is wrong, email info@cinderandcrown.com with the number, where you saw it, and the official source. We check it against the source, fix confirmed errors and note the fix in the site's change log.