Verstappen 2026: the fight for second
Three million simulated seasons. Six races remaining. Explore where Max could finish, why second leads the distribution, and what the title long shot really means.
Turkish PDF · preserved as supplied
Simulated seasons
The main forecast sample.
Chance of finishing P2
The most likely single finishing position.
Expected season points
The model’s mean total.
Executive summary
Second is the most likely single outcome
In the supplied analysis, Max Verstappen finishes second in 35.04% of the three-million-season main simulation. The combined top-three probability is 58.37%. No single finishing position exceeds 50%.
A title is possible, but highly unlikely
The separate importance-sampling estimate is approximately 0.0190%. It is conditional on the model and its assumptions, not a calibrated real-world guarantee.
Read the assumptions alongside the numbers
The analysis uses a 16 October 2026 data cutoff, after Singapore. It assumes six remaining full-distance Grands Prix, no remaining sprints and the same 22-driver race lineup. This English web edition presents the supplied PDF. The calculation files referenced in the PDF were not supplied and have not been rerun here.
This is an English presentation of the supplied report dated 16 October 2026. Its estimates are conditional on that report’s data and assumptions. More simulation runs reduce counting noise; they do not remove uncertainty about car performance, upgrades or reliability.
Second is the strongest signal.
No single finishing position exceeds 50% in the combined forecast. Change the model to see how the weight placed on recent form changes the picture.
Hover, focus or tap a row for details. Press Escape to close.
In the combined sample, P8 remains possible despite zero observed finishes. P9–P23 are mathematically impossible under the report’s fixed-points and six-race assumptions. The report does not give model-specific P7–P8 percentages; striped tracks mean “not reported,” not zero. Rounded values are not renormalized. Very small bars may be invisible at this scale; the labels retain the published precision.
These three figures always refer to the equal-weight combined model, regardless of the chart selection.
See exact counts and sampling uncertainty +
All rows below refer to the main 3,000,000-season combined sample. The 95% intervals describe counting uncertainty within the fitted model. They do not measure real-world forecast error.
| Finish | Standard MC | Count / 3,000,000 | 95% counting interval / bound |
|---|---|---|---|
| P1 | 0.0188% | 564 | 0.0173–0.0204% |
| P2 | 35.04% | 1,051,132 | 34.98–35.09% |
| P3 | 23.31% | 699,373 | 23.26–23.36% |
| P4 | 19.35% | 580,493 | 19.31–19.39% |
| P5 | 15.69% | 470,849 | 15.65–15.74% |
| P6 | 6.58% | 197,423 | 6.55–6.61% |
| P7 | 0.0055% | 166 | 0.0048–0.0064% |
| P8 | Not observed | 0 | 95% one-sided upper bound: 0.00010% |
| P9–P23 | 0% | Mathematically impossible | With the stated starting points and schedule |
The separate championship importance-sampling estimate is 0.0190%. It has not been substituted into, or used to renormalize, this standard Monte Carlo distribution.
A range, not a promise.
The report’s mean is 299.8 points; its median is 299. The 90% model interval is 258–342, and the 80% interval is 267–333.
Intervals share a 240–360 point scale. Median: 299.
Hover, focus or tap a band or the mean for details.
Explore three additional sensitivity checks +
| Model | P1 | P2 | P3 | P4 | P5 | P6 |
|---|---|---|---|---|---|---|
| Higher uncertainty | 0.0117% | 34.1% | 24.8% | 19.9% | 15.5% | 5.8% |
| Track effect off | 0.0124% | 37.9% | 24.6% | 18.7% | 14.0% | 4.7% |
| Fixed parameters | 0.0010% | 40.3% | 25.1% | 18.3% | 13.2% | 3.1% |
Higher uncertainty: Balanced model with race-performance spread increased by 20% and persistent strength uncertainty by 25%.
Track effect off: Historical track and track-similarity adjustments set to zero.
Fixed parameters: Strength and DNF parameters held fixed; race randomness remains. This is a sensitivity check, not a claim that parameters are known.
One driver. A whole field of rivals.
The report’s snapshot covers 17 Grands Prix and six sprints: 23 championship entries, with 22 starters assumed for the remaining races. These are the values in the supplied document, not live championship standings.
Scroll the table horizontally to see every metric.
| Pos. | Driver / team | Points | Last 5 GPs | DNF / GP | Projected mean |
|---|---|---|---|---|---|
| 01 | Kimi AntonelliMercedes | 340 | 93 | 1/17 | 415.7 |
| 02 | George RussellMercedes | 238 | 55 | 3/17 | 293.2 |
| 03 | Lewis HamiltonFerrari | 229 | 39 | 1/17 | 296.9 |
| 04 | Max VerstappenRed Bull | 221 | 101 | 4/17 | 299.8 |
| 05 | Charles LeclercFerrari | 215 | 54 | 3/17 | 286.0 |
| 06 | Lando NorrisMcLaren | 201 | 41 | 4/17 | 254.7 |
| 07 | Oscar PiastriMcLaren | 138 | 32 | 3/17 | 175.9 |
| 08 | Isack HadjarRed Bull | 97 | 26 | 2/14 | 139.2 |
| 09 | Liam LawsonRacing Bulls | 75 | 20 | 1/17 | 104.0 |
Last-five figures contain Grand Prix points only; season totals include sprints. “DNF” also covers DNS, DSQ and other non-normal finishes. Projected means do not directly determine the most likely championship order.
Driver changes and snapshot assumptions +
* Yuki Tsunoda is listed as a substitute in three Grands Prix in the report, with one point retained. The remaining races use the report’s final 22-driver grid. Hadjar’s three-race absence is not counted as a DNF. Lawson’s Red Bull and Racing Bulls spells were fitted as separate driver/car combinations.
The report states that its reconstructed totals matched all 23 drivers in the official table. This presentation transcribes that claim and the table; it does not independently reconstruct the underlying race archive.
150 points left on the table.
Six full-distance Grands Prix, no remaining sprints and no fastest-lap bonus are assumed in the report. Each race below shows Max’s conditional win probability and expected additional points.
Austin
USAFast esses. A mixed-speed challenge.
Mexico City
MEXHigh altitude. More energy and cooling uncertainty.
São Paulo
BRAElevation, varied corners and strategy variability.
Las Vegas
USAA fast street circuit. Low drag matters.
Lusail
QATFlowing medium- and high-speed corners.
Abu Dhabi
UAEMixed speeds and traction zones.
Dates follow the report’s Türkiye calendar days. Las Vegas is 21 November locally and 22 November in Türkiye. Historical circuit effects are heavily reduced for the 2026 technical change. Track groupings are modeling assumptions, not measured vehicle physics; no reliable circuit weather forecast is implied.
A title is possible. The path is tiny.
Approximately one in 5,275 similar modeled seasons.
Model dependent. Standard Monte Carlo: 0.0188%. These are two distinct estimates.Max needs help from the other side.
Starting on 221 points, Max can reach 371 with six wins. Antonelli starts on 340. An additional 30 points leaves Antonelli behind; 31 produces a tie that must be resolved by countback.
Five wins and a second place put Max on 364. Antonelli must then score at most 23 more: at 364–364, the report’s existing wins give Antonelli the tiebreak.
A strong finish from Max alone is not enough. This tail depends heavily on an unusually poor closing sequence for Antonelli.
Inspect the rare-event estimates +
Importance sampling draws title-enabling scenarios more often, then corrects their contribution using the normal-model density divided by the proposal density. Proposal event counts alone are not probabilities.
| Model | Titles in 1M standard seasons | Importance estimate | Approx. 95% sampling interval | Effective sample |
|---|---|---|---|---|
| Season-wide | 5 | 0.000548% | 0.000329–0.000767% | 24 |
| Balanced form | 109 | 0.0110% | 0.0092–0.0129% | 132 |
| Recent form | 450 | 0.0453% | 0.0426–0.0479% | 1,141 |
| Higher uncertainty | 117 | 0.0126% | 0.0116–0.0137% | 552 |
The combined estimate’s numerical standard error is 0.000551 percentage points; it excludes model error. The report describes replacing an unstable exploratory proposal with Gaussian shifts in persistent strength and smaller Gumbel shifts. Low effective sample sizes still limit precision.
Zero counts ≠ zero chance.
No P8 finishes appeared in the main three million seasons. The resulting one-sided 95% upper bound is 0.00010%, conditional on the model. A separate importance-sampling attempt was too unstable to justify a point estimate.
A valid path to eighth +
Across all six races, let Hadjar finish first, Piastri second, Norris third, Leclerc fourth, Hamilton fifth, Russell sixth and Antonelli seventh, while Max scores nothing. Seven drivers then finish ahead of him on points.
The floor has a proof.
Gasly and the drivers below him cannot reach Max’s existing 221 points with 150 available. Lawson needs all six wins to pass Max; that leaves Hadjar at most 108 extra points, or 205 total. Both cannot pass Max together, so at most seven rivals can finish ahead.
Why the P8 tail estimate was withheld +
The separate proposal produced 0, 8 and 25 P8 events across the three base models, but effective samples were only 0.0, 1.1 and 1.7. The largest weight supplied 95.5% of the balanced estimate and 76.6% of the recent-form estimate. The report therefore reports no reliable P8 point estimate.
What the model knows. What it doesn’t.
A season-level ranking model sits behind these figures. It represents race outcomes and shared uncertainty, rather than simulating every lap of a car’s physics.
Results, qualifying and sprints.
The report uses 17 Grands Prix, 17 qualifying sessions and six sprints from 2026, plus 48 Grands Prix from 2024–2025. Race, qualifying and sprint contributions are weighted 1, 0.30 and 0.25.
A shared race for 22 drivers.
A Plackett–Luce ranking model fits driver/car combinations using the relative order of normal finishers, with weak teammate regularization. A retirement is not treated as proof of slow pace. Every race has one joint finishing order.
Form can persist across races.
Strength is sampled for each season using an approximate fitted covariance. Team setup shocks and individual Gumbel performance variation then act in each race. Past improvement is not mechanically extended into the future.
Points and countback matter.
The model applies 25–18–15–12–10–8–6–4–2–1 points and existing sprint points. Ties compare Grand Prix wins, then seconds and subsequent finishes. The report says all ties involving Max resolved at that stage.
The main 3M sample is part of the 6M total standard computation, not additional to it. Backtesting and development runs are excluded from these final-computation counts.
Reliability: why raw DNF rates are adjusted +
Reliability estimates use an eight-Grand-Prix time-weighting half-life and shrink driver/car evidence toward team and grid averages. A beta draw represents persistent uncertainty in each driver’s DNF probability. Shared race and team risks use a Gaussian copula with assumed shares of 0.08 and 0.12; the report does not claim the small dataset identifies these values precisely.
| Driver | DNF / GP | Raw rate | Model mean per race |
|---|---|---|---|
| Kimi Antonelli | 1/17 | 5.9% | 10.4% |
| George Russell | 3/17 | 17.6% | 16.9% |
| Lewis Hamilton | 1/17 | 5.9% | 12.1% |
| Max Verstappen | 4/17 | 23.5% | 17.2% |
| Charles Leclerc | 3/17 | 17.6% | 15.6% |
| Lando Norris | 4/17 | 23.5% | 18.6% |
| Oscar Piastri | 3/17 | 17.6% | 15.1% |
| Isack Hadjar | 2/14 | 14.3% | 12.6% |
| Liam Lawson | 1/17 | 5.9% | 7.6% |
A useful check. A small sample.
From Hungary through Singapore, each target used only earlier race results: 30,000 draws per model and race, or 630,000 single-race simulations in total. Lower scores are better.
The season-wide model was narrowly ahead. Seven races are insufficient to establish a decisive model choice or real-world calibration.
| Model | Winner Brier ↓ | Podium Brier ↓ | Points MAE ↓ |
|---|---|---|---|
| Balanced form | 0.8446 | 0.0865 | 3.193 |
| Recent form | 0.8409 | 0.0892 | 3.241 |
| Season-wide | 0.8403 | 0.0824 | 3.118 |
The backtest uses the final archive available at the report’s cutoff. It does not reconstruct the provisional, pre-appeal versions of results that were available on each historical day.
The unknowns stay unknown.
- Fuel, energy deployment, tyre temperatures, individual pit stops and safety-car decisions are not simulated lap by lap. Race-level shocks stand in for their combined effect.
- Future engine penalties, upgrades, injuries and team orders are not forecast as separate event processes.
- Full-distance races and the six-event schedule are assumed. Cancellations or shortened races require a fresh calculation.
- The report approximates retirement classification using completed-distance fractions. Detailed lap-based classification, including lapped finishers, is not reproduced. This limits realism in extreme mass-retirement scenarios.
- The 2026 technical change limits the relevance of older seasons. Historic track effects are small adjustments, not the main source of pace estimates.
Source material and reproducibility status +
This web edition is a translation and interactive presentation of the supplied nine-page PDF. Figures, assumptions, reported checks and conclusions are attributed to that document. It is not an official F1 or FIA forecast.
The PDF cites official driver standings, 2024–2026 result indexes, the remaining race and sprint calendars, FIA Sections A and B, and technical/team reporting. Its full bibliography is on page 9.
The document describes a source manifest with 88 downloaded result tables, raw-data files and Python scripts including simulate.py and refine_rare.py. Those files and their random seeds were not supplied with this PDF. The simulations have therefore not been independently rerun for this web edition.
The page’s controls compare published outputs only. Future reports can be added as new dated editions, preserving the assumptions and data cutoff behind each forecast.
Second place is a serious possibility.
The championship remains an extreme long shot.