Austin 2026: the race forecast
Three million simulated races. Twenty-two drivers. Explore the United States Grand Prix forecast, all result-order probabilities and nine views of the assumptions behind them.
Executive summary
Verstappen leads, but the race is open
Max Verstappen has the highest win probability in the main forecast at 39.22%, followed by Charles Leclerc at 16.40% and George Russell at 13.95%. A favourite losing is entirely consistent with these probabilities. The central order in the PDF is sorted by expected position; it is not a predicted starting grid or a win-probability ranking.
A forecast made before the weekend
The supplied report uses a 20 October 2026 data cutoff, before Austin practice and qualifying. The 25 October Grand Prix is modelled over 56 laps. The starting grid is simulated, not observed. No driver-specific reputation bonus is applied.
The assumptions can move the result
In the fully wet scenario, Verstappen's win probability falls to 30.79%. The hypothetical 0.15-second-per-lap McLaren improvement raises Norris to 8.84% and Piastri to 1.53%. These are conditional scenarios, not forecasts that rain or an upgrade will occur.
Read the full order carefully
P1–P22 covers every entrant, including retired, non-starting and disqualified cars. It is not equivalent to FIA classification. Points require a classified top-ten result; DNF/DNS and DSQ indicators can overlap and should not be added.
The model has not been rerun for this website
The English web edition uses the exact numerical data supplied in the interactive HTML and checks the main forecast against the supplied CSV. The supplied 12-page English PDF, CSV and interactive HTML are available below. The source code and full reproduction package referenced in the PDF were not supplied.
One race. Nine perspectives.
Explore the stored results of the main forecast, three form models and five conditional scenarios.
Equal mixture of three form models with 2-, 4- and 8-race half-lives. Each simulated race has one shared 22-driver field.
- Simulated races
- 3,000,000Selected scenario sample
- Wet race
- 5.00%Realised model frequency
- Safety Car
- 29.65%Realised model frequency
- Virtual Safety Car
- 30.50%Realised model frequency
The main forecast pools 3 million form-model runs; each other view contains 1 million runs. The individual form views reuse those pooled runs. Scenario changes do not run a new simulation. Event frequencies can overlap.
Who leads this scenario?
The six highest win estimates, on the same 0–50% scale in every view. Hover, focus or tap a row for its observation count.
The outcome probabilities.
Scoring probability is supplied by the model and requires an eligible, classified top-ten result. It is not simply the sum of the first ten result-order slots.
22 of 22 driversSelect a column heading to sort.
| Max VerstappenVER · Red Bull Racing | 39.22% | 68.09% | 81.92% | 17.68% | 0.46% | 5.22 | 16.09 |
|---|---|---|---|---|---|---|---|
| Charles LeclercLEC · Ferrari | 16.40% | 51.41% | 83.33% | 16.13% | 0.47% | 5.85 | 13.02 |
| George RussellRUS · Mercedes | 13.95% | 44.11% | 81.71% | 17.48% | 0.47% | 6.40 | 11.95 |
| Kimi AntonelliANT · Mercedes | 12.68% | 46.17% | 88.41% | 10.84% | 0.46% | 5.40 | 12.66 |
| Lewis HamiltonHAM · Ferrari | 7.13% | 34.62% | 86.35% | 12.60% | 0.47% | 6.19 | 10.93 |
| Isack HadjarHAD · Red Bull Racing | 5.17% | 22.79% | 83.63% | 13.13% | 0.46% | 7.24 | 8.77 |
| Lando NorrisNOR · McLaren | 4.44% | 21.73% | 78.75% | 18.99% | 0.46% | 7.84 | 8.54 |
| Oscar PiastriPIA · McLaren | 0.63% | 6.16% | 78.19% | 15.52% | 0.46% | 8.85 | 5.66 |
| Liam LawsonLAW · Racing Bulls | 0.25% | 2.90% | 77.23% | 8.30% | 0.46% | 9.01 | 4.38 |
| Nico HulkenbergHUL · Audi | 0.04% | 0.63% | 52.23% | 16.93% | 0.46% | 11.41 | 2.14 |
| Franco ColapintoCOL · Alpine | 0.04% | 0.44% | 34.94% | 19.02% | 0.47% | 12.71 | 1.30 |
| Pierre GaslyGAS · Alpine | 0.02% | 0.34% | 44.34% | 15.83% | 0.47% | 11.89 | 1.62 |
| Arvid LindbladLIN · Racing Bulls | <0.01% | 0.19% | 38.73% | 7.56% | 0.46% | 11.89 | 1.28 |
| Oliver BearmanBEA · Haas F1 Team | <0.01% | 0.15% | 27.22% | 17.47% | 0.47% | 13.18 | 0.88 |
| Lance StrollSTR · Aston Martin | <0.01% | 0.08% | 6.45% | 53.49% | 0.47% | 17.32 | 0.22 |
| Esteban OconOCO · Haas F1 Team | <0.01% | 0.07% | 19.33% | 12.21% | 0.46% | 13.58 | 0.56 |
| Gabriel BortoletoBOR · Audi | <0.01% | 0.05% | 16.93% | 15.89% | 0.46% | 14.08 | 0.47 |
| Alexander AlbonALB · Williams | <0.01% | 0.02% | 4.83% | 40.40% | 0.47% | 16.85 | 0.13 |
| Carlos SainzSAI · Williams | <0.01% | 0.02% | 9.34% | 24.58% | 0.46% | 15.32 | 0.25 |
| Fernando AlonsoALO · Aston Martin | <0.01% | <0.01% | 4.18% | 35.83% | 0.46% | 16.74 | 0.11 |
| Sergio PerezPER · Cadillac | <0.01% | <0.01% | 1.12% | 31.15% | 0.46% | 17.61 | 0.02 |
| Valtteri BottasBOT · Cadillac | <0.01% | <0.01% | 0.51% | 39.60% | 0.46% | 18.40 | 0.01 |
Read the flags separately. DNF/DNS means did not finish or did not start; DSQ means disqualified. These flags can overlap each other and the result-order slots. Do not add them together. Mean order includes all 22 participants; expected points cover the Sunday Grand Prix only.
0* means no observations in the selected sample, not impossibility. <0.01% is a non-zero estimate below the displayed precision. The CSV exports all 22 drivers regardless of search, with raw probabilities from 0 to 1 and exact result-order counts.
Every driver. Every position.
This is the model’s full P1–P22 result order, not FIA classification. NC, DNF, DNS and DSQ participants are included. Retirees are ordered by laps and time; disqualifications are placed at the end. A DNF can still be FIA-classified after reaching the required 90% lap threshold.
Focus a cell, then use the arrow keys to explore. Enter pins the details; Escape closes them. Scroll to see every position. Driver order stays fixed between scenarios.
| Driver / result order | P1 | P2 | P3 | P4 | P5 | P6 | P7 | P8 | P9 | P10 | P11 | P12 | P13 | P14 | P15 | P16 | P17 | P18 | P19 | P20 | P21 | P22 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Max VerstappenVER | ||||||||||||||||||||||
| Charles LeclercLEC | ||||||||||||||||||||||
| Kimi AntonelliANT | ||||||||||||||||||||||
| Lando NorrisNOR | ||||||||||||||||||||||
| Oscar PiastriPIA | ||||||||||||||||||||||
| Lewis HamiltonHAM | ||||||||||||||||||||||
| Liam LawsonLAW | ||||||||||||||||||||||
| Nico HulkenbergHUL | ||||||||||||||||||||||
| George RussellRUS | ||||||||||||||||||||||
| Isack HadjarHAD | ||||||||||||||||||||||
| Pierre GaslyGAS | ||||||||||||||||||||||
| Oliver BearmanBEA | ||||||||||||||||||||||
| Esteban OconOCO | ||||||||||||||||||||||
| Carlos SainzSAI | ||||||||||||||||||||||
| Arvid LindbladLIN | ||||||||||||||||||||||
| Fernando AlonsoALO | ||||||||||||||||||||||
| Sergio PerezPER | ||||||||||||||||||||||
| Alexander AlbonALB | ||||||||||||||||||||||
| Valtteri BottasBOT | ||||||||||||||||||||||
| Lance StrollSTR | ||||||||||||||||||||||
| Franco ColapintoCOL | ||||||||||||||||||||||
| Gabriel BortoletoBOR |
Each row and column totals 100% before rounding. Colour intensity is fixed across scenarios and saturates at 50%. For zero observations, the approximate one-sided 95% Monte Carlo upper bound is 0.000100%, assuming independent repeats of this model. That counting bound does not include model error.
Follow one driver.
- Mean result order
- 5.22
- Expected Sunday points
- 16.09
- DNF / DNS
- 17.68%
- DSQ
- 0.46%
Explore Max Verstappen’s 22-position distribution
Full result-order positions, including non-finishers and disqualifications. Every bar uses a 0–100% scale.
The assumptions are part of the forecast.
English editorial summary of the supplied PDF, pages 4–9 and 12. The displayed simulations and validation results were reported by the source author; they have not been rerun here.
One shared race, 22 entrants.
The main forecast pools one million races from each of three models. Recent form decays with a half-life of two, four or eight races, with equal mixture weights. All drivers follow the same modelling rules.
Race and qualifying order are learned separately. Sprint observations receive 0.30 of a Grand Prix observation's weight. The simulated grid feeds a 56-lap race with tyre wear, pit stops, overtaking, reliability, incidents and penalties.
Eight million distinct forecast races.
The three-million-race main forecast plus five one-million-race stress scenarios make eight million. The H2, H4 and H8 views show the components already pooled into the main forecast; they are not extra runs.
The PDF separately reports 80,000 calibration and 420,000 validation races. These are excluded from the forecast sample.
A dated set of assumptions.
The 5% wet-race prior and 30% Safety Car / 30% VSC targets are model assumptions. The main pit-loss estimate is 20.6 seconds. The McLaren scenario is a hypothetical equal pace gain for both cars.
There is at most one Safety Car and one VSC phase. Detailed changing rain, energy management, repeated restarts, cancelled races and reduced-points events are not fully modelled.
Seven test races, limited evidence.
On the reported seven-race evaluation, the main model's favourite won three times. Its rank score was 0.1117 and mean absolute rank error was 3.42 positions. The simple baseline performed better on podium and points Brier scores.
This was a retrospective evaluation, not a blind prospective trial. The small differences do not establish reliable superiority or high predictive accuracy.
Source provenance & reproducibility
The supplied package contains a 12-page PDF, a main-forecast CSV and an HTML file with nine scenario datasets. The full model code, downloaded evidence archive and run metadata referenced on PDF page 12 were not supplied, so this website presents the reported results rather than claiming to reproduce the engine.
The CSV's 616 numeric values match the main HTML dataset exactly. Every driver's 22-position distribution sums to 100%, and every position has one entrant per simulated race. These are data-integrity checks, not proof that the model predicts reality accurately.
Data cutoff: 20 October 2026. The source does not specify a cutoff time. Event date checked against the official Formula 1 schedule on 21 October 2026.
- Forecast reportSHA-256:
597c0e88b0891494774aa55928d35878f85429f5e98edb7b8ad21f0a3d7d37bd - Full probability datasetSHA-256:
1f25b63f18eb56cb746bc9ca5ebce97de9bf655d9a282cca602e34a014818db1 - Interactive resultsSHA-256:
36d0792a8f87dbffbd9a943629a40ff2b5c694def0d194ded242556d9f60bff9
Keep the evidence close.
Download the supplied English editions, preserved byte for byte. The CSV and offline interactive HTML contain the same forecast data used in the charts above. The full 12-page English PDF is also available below.