2026 Presidential Model
Win probability
61%
Flávio
39%
Lula
In every 100 simulated elections, Flávio wins 61 and Lula wins 39.
Simulations
- Cury<1%
- Zema<1%
- Renan<1%
- Caiado<1%
Each square represents one simulated election, colored by the winner of that simulation. Rounded to 100 simulations.
First round
Simulated projections for election day, as a share of valid votes. The shaded band covers 90% of simulations; the line is the median.
The election is decided in the first round in <1% of simulations.
median90% band
Where each candidate finishes
How often each candidate finishes in each position in the first round (not to be confused with winning in the first round)
| Candidate | 1st | 2nd | 3rd | 4th | 5th |
|---|---|---|---|---|---|
| LulaLulaPT | 66% | 34% | <1% | <1% | <1% |
| Flávio BolsonaroFlávioPL | 34% | 66% | <1% | <1% | <1% |
| Augusto CuryCuryAVANTE | <1% | <1% | 87% | 11% | 1% |
| Renan SantosRenanMISSÃO | <1% | <1% | 8% | 60% | 31% |
| Romeu ZemaZemaNOVO | <1% | <1% | <1% | <1% | 4% |
| Ronaldo CaiadoCaiadoPSD | <1% | <1% | 4% | 28% | 64% |
The runoff
Projected runoff vote as a share of valid votes.
The 50% line is the one that decides the runoff.
| Opponent | Runoff chance | Lula's vote | Lula's win |
|---|---|---|---|
| Flávio BolsonaroFlávioPL | 99% | 49.3% | 39% |
| Augusto CuryCuryAVANTE | <1% | 49.1% | 39% |
| Renan SantosRenanMISSÃO | <1% | 51.0% | 57% |
| Romeu ZemaZemaNOVO | <1% | 48.6% | 11% |
| Ronaldo CaiadoCaiadoPSD | <1% | 49.1% | 67% |
How often each result comes up
Distribution of simulations by first-round vote band. Hover a band to read its value.
The forecast over time RECONSTRUCTED
The charts below show a reconstruction of the model from the polls released each day since 5 April. The model was published on 15 August.
Latest note: 14 Sep — How to evaluate the presidential model's performance
Each candidate's win probability, stacked -- each band is that candidate's slice of the day's 100%. Hover over the chart to see the numbers for a specific day.
The same series, one line per candidate, on the same 0-100% scale. Hover over the chart to see the numbers for a specific day.
The model used 65 polls on 5 April 2026; it uses 439 today.
How the model works
The model simulates the election tens of thousands of times. Each simulation starts from the aggregated poll average, adjusts uncertainty using the error polls have historically shown at this point in the electoral calendar, allows for the movement that can still occur before voting day, and resolves the runoff from projections of how votes transferred in past elections. See the methodology.
What this model is not
- This is not a model that projects results state by state.
- It does not assume polls always err in the same direction.
- It is not an election poll.