Texas · U.S. House · District 33 · November 3, 2026

What it would take for Patrick Gillespie.

Allred starts 38.5 points ahead on our starting line from recent official results. Campaign work can move about 3,566 net votes toward Gillespie, a 2.4-point gain. Winning outright would take a 36.1-point national wave on top of that. Here is the plan that makes Gillespie the one standing if one comes.

57,747

net votes between Gillespie and a win

days to election day. Early voting opens October 19.

38.5 pts

Allred lead, our starting line from recent results

In short

One-page brief ›All Texas races ranked ›

A long shot on the math: it takes a 36.1-point national wave.

Odds today0%

Odds with the plan0%

Net votes to find57,747

WherePrecinct 1060 (Dallas County), Precinct 2605 (Dallas County), Precinct 4079 (Dallas County)

Underdog UpdateSeptember 26, 2026. Underdog plan published. Starting line: Allred +38.5. Patrick Gillespie needs 57,747 net votes. Odds 0% today, 0% with the plan.

The number

74,955 votes.

That is half of the 149,908 ballots we expect in this race, plus one. Where Gillespie stands today, and what is missing:

44,119 Patrick Gillespie baseline

101,866 Colin Allred baseline

Starting line: a weighted blend of the official 2022 and 2024 results in this district (President, U.S. Senate, Railroad Commissioner, Governor, Attorney General). No 2026 polls have been published. 2026 turnout modeled on 2022, the last midterm.

The odds

0% today. A wave changes that.

We ran this race 20,000 times with realistic error on every assumption. Drag the slider to see the odds move as the campaign captures more of what history says is available.

Gillespie wins in

of 20,000 simulated elections

With no plan

How much of what history allows the campaign achieves100%

The plan on this page sits at 100%. Median outcome at this setting: Gillespie short by 54,315.

Each simulation draws the starting line with an error of 7 points of margin (no 2026 polls, so the starting line comes from history and carries more error) and lets the plan over- or under-deliver by 30 percent. Seeded and reproducible; the method and data are at the bottom of the page.

Three moves

The gap closes three ways.

Change minds. Bring out voters who skip midterms. Win back the third-party vote. Every pool below is counted from official precinct returns and published polls.

Persuade

3,128

net votes in the plan, from a pool of 3,128

Who. 3,128 voters who split their ticket in recent elections and lean Allred today.

How. Live conversations on the call list that matters: voters who crossed party lines, plus the undecided. Gillespie needs to win 50 percent of the ones now leaning Allred.

Mobilize

net votes in the plan, from a pool of 110

Who. At least 110 people who voted in 2024 but not in 2022, in precincts where Gillespie's side wins at least 55 percent.

How. Voter ID calls to people who voted in 2024 and skipped 2022, then a ballot chase until each one has voted.

Consolidate

415

net votes in the plan, from a pool of 1,186

Who. Voters who went third party in recent races, in the precincts that lean Gillespie's way.

How. A direct appeal to voters who went third party last time. They have already shown they will leave the two-party default.

Together: 3,566 net votes against a gap of 57,747. Still 54,181 short. The rest needs a national shift of about 36.1 points that no campaign controls.

Where it is won

12 precincts carry 18% of it.

Ranked by the net votes each one gives the plan. The darker the county, the more of the upset lives there.

Fewer net votesMore net votes for the upset

01
Precinct 1060 (Dallas County)persuade
Starting line Allred +33. 5.0% split their ticket.
+84
02
Precinct 2605 (Dallas County)persuade
Starting line Allred +28. 7.6% split their ticket.
+58
03
Precinct 4079 (Dallas County)persuade
Starting line Allred +51. 4.1% split their ticket.
+57
04
Precinct 2602 (Dallas County)persuade
Starting line Allred +27. 6.0% split their ticket.
+54
05
Precinct 2102 (Dallas County)persuade
Starting line Allred +15. 4.1% split their ticket.
+51
06
Precinct 3000 (Dallas County)persuade
Starting line Allred +14. 4.4% split their ticket.
+51
07
Precinct 1048 (Dallas County)persuade
Starting line Allred +20. 4.6% split their ticket.
+50
08
Precinct 2603 (Dallas County)persuade
Starting line Allred +26. 5.7% split their ticket.
+48
09
Precinct 2147 (Dallas County)persuade
Starting line Allred +25. 6.6% split their ticket.
+48
10
Precinct 2604 (Dallas County)persuade
Starting line Allred +30. 4.9% split their ticket.
+44
11
Precinct 4521 (Dallas County)persuade
Starting line Allred +15. 3.5% split their ticket.
+44
12
Precinct 2099 (Dallas County)persuade
Starting line Allred +44. 3.7% split their ticket.
+43

What the numbers say that the polls do not

The hidden path.

Precinct 1060 (Dallas County) carries the most.

Precinct 1060 (Dallas County) gives this plan more net votes than any other precinct: 84. It sits at Allred +33 on our starting line, and 5.0 percent of its voters split their ticket in recent elections.

Precinct 2605 (Dallas County) splits its ticket the most.

Of the precincts in this plan, Precinct 2605 (Dallas County) has the highest share of ticket splitters: 7.6 percent. That is where a persuasion call is most likely to reach someone open to crossing over.

The honest math.

Every lever at its historical ceiling moves 3,566 net votes. The gap is 57,747. That is the measure of what a strong campaign adds: 2.4 points of margin, and first position if the national mood breaks.

The evidence

Where voters cross over.

Every dot is a precinct. If no one split a ticket, every dot would sit on the line. The orange ones are where Gillespie's side lost voters it won further up the ballot.

Precincts: 258
Slope: 0.976
R²: 0.958
Residual SD: 3.5 pts
Crossover precincts: 63

Weighted least squares, one point per precinct, weighted by ballots cast. Orange precincts sit more than one residual standard deviation below the line: Gillespie's side did worse there in this race than the top of the ticket predicts. Hover any dot for the precinct.

The next 37 days

Voting starts October 19.

Now to October 5
Find them
Voter ID calls across the persuasion list in these precincts. Every conversation tags the voter: with Gillespie, with Allred, or undecided. The undecided become the persuasion universe; the supporters become the ballot chase list.
October 6 to 18
Persuade
Second conversations with every undecided and soft voter, built around the one issue they named on the first call. The goal is 3,128 net votes from people who have crossed party lines before.
October 19 to 25
Bank the ballots
Mail ballots are already going out, and in-person early voting runs October 19 through October 30. Mail ballot applications must be received by October 23. Chase every supporter's mail ballot request until it is returned, and move supporters to early voting while it is open.
October 26 to November 2
Turn them out
The drop-off voters: people who voted in 2024 and skipped 2022. At least 110 of them live where Gillespie is favored. In-person early voting ends October 30. Mail ballots count if received by 7 p.m. on November 3, or by 5 p.m. on November 4 if postmarked by Election Day.
November 3
Election day
Polls are open 7 a.m. to 7 p.m. Texas has no same-day registration; the registration deadline is October 5. The last calls go to identified supporters who have not voted, with their polling place in hand.

What has to be true

No wishful thinking.

Official results in this district: Harris carried the 2024 presidential race by 32.7 points; Allred carried the 2024 U.S. Senate race by 40.1 points; O'Rourke carried the 2022 governor's race by 44.0 points.

Gillespie wins 50 percent of the ticket splitters now leaning Allred.
The national environment moves about 36.1 points toward Gillespie's side between now and November 3.
No 2026 public poll exists for this race, so the starting line is a weighted blend of official recent results; the first real poll will move it.

The work

819 conversations.

That is how many live voter conversations this plan takes: about 15,633 dials to 5,211 households, before the last ballot is cast.

5,211

households on the target list

16K

dials across three attempts

819

live conversations: persuasion, voter ID, ballot chase

Call list: every ticket splitter and undecided voter in the targeted places (you cannot tell which way they lean before the call), every targeted drop-off voter, and the reachable third-party voters. About 69 households for every 100 voters on the list, three attempts each, and a 5.24 percent live answer rate per attempt, the per-program average across 1.2 million real 2026 dials in our published benchmarks.

HyperPhonebank

We run this list.

Calls are the voter ID and ballot chase engine of this plan. They find the persuadable voters and the supporters, so every other dollar lands on the right door. The full call program for this list is an estimated $2,714 at our published rates: 5,211 households, three attempts each.

Professional live callers working exactly these precincts: persuasion on the ticket splitters, voter ID on the fence, ballot chase on every absentee request until it is returned. A live dashboard shows every conversation as it happens. Days to launch, not weeks.

Run this plan How HyperPhonebank works ›Size your own program ›

The candidates

Who is on the ballot.

Underdog

Patrick Gillespie (R)

Customs broker and former technical writer who worked for Boeing and BAE Systems, and a former vice president of International Association of Machinists Local Lodge 2768.

Campaign site ›FEC filings ›

Favorite

Colin Allred (D)

Former U.S. representative for the 32nd District (2019 to 2025), civil rights attorney and former Tennessee Titans linebacker; a Baylor and UC Berkeley law graduate who served at HUD in the Obama administration.

Cash on hand $159,887 and $7,431,941 raised since January 1, 2025, per FEC reports through June 30, 2026.

FEC filings ›Wikipedia ›

Underdog Updates

What changed, and when.

We re-run this race as new polls, fundraising reports and early vote counts arrive. Every change to the numbers is logged here.

September 26, 2026
Underdog plan published
Starting line: Allred +38.5. Patrick Gillespie needs 57,747 net votes. Odds 0% today, 0% with the plan.

Questions

Can Patrick Gillespie win Texas's 33rd Congressional District in 2026?

Only with help from the national environment. Campaign work can move about 3,566 net votes; the gap is 57,747. The rest would take a 36.1-point national shift.

Where are the votes Patrick Gillespie needs?

Ranked by net votes in the plan: Precinct 1060 (Dallas County), Precinct 2605 (Dallas County), Precinct 4079 (Dallas County), Precinct 2602 (Dallas County), Precinct 2102 (Dallas County). Persuadable voters are counted from ticket splitting in 2022 and 2024 official returns; turnout targets are people who voted in 2024 and skipped 2022.

How many voters does Patrick Gillespie need to persuade?

About 1,564 voters who lean Colin Allred today, each worth two votes of margin when they switch. That is 50 percent of the 3,128 persuadable voters leaning Colin Allred.

When does early voting start in Texas in 2026?

In-person early voting in Texas runs October 19 through October 30, 2026. The voter registration deadline is October 5, and applications for a mail ballot must be received by October 23. Mail ballots count if received by 7 p.m. on Election Day, or by 5 p.m. the next day if postmarked by Election Day. Texas does not offer same-day registration.

Who made this analysis?

The Political Group, an independent campaign services firm, from official Texas Secretary of State precinct returns, Census data, FEC filings and published polls. It is not affiliated with any candidate or party.

Method and sources

Precinct math from official Texas Secretary of State returns for 2022 and 2024, projected onto current district lines. Polls, ratings and fundraising as published on the dates shown. This is independent analysis by The Political Group, as of September 26, 2026. It is not affiliated with or endorsed by any candidate or party, and it is written for every campaign that starts behind.

Statistics: a weighted least squares crossover regression on every precinct, and 20,000 seeded Monte Carlo simulations. The baseline error (7 points of margin) and plan efficacy spread (30 percent) are our stated assumptions. Download the precinct data (CSV, ready for JASP). Free to cite under CC BY 4.0 with a link to this page.