Simulation desk
Simulation desk
HowTap a driver, then tap a seat. Fill all 22 and run it.Drag a driver into a seat, or tap then tap. Fill all 22 and run it.
Twenty-two seats. Move anybody anywhere, pull a prospect up off the bench, send a driver home, then run ten thousand seasons on the grid you built.
Reserves, juniors and the F2 front row. Nobody here has a seat. Put one in a car and the model prices him with a wide band, because nobody knows yet.
Nobody has driven the 2027 cars. Every question here is one the paddock is actually asking, with the reporting that makes it a live question under it. Anything that could not be sourced is not a lever.
Two numbers, not one. A car rating belongs to the seat and a skill rating belongs to the driver, both on a 0 to 100 scale where a point is worth roughly a tenth and a half over a lap. Race pace is car plus skill plus the weekend. Move a driver and he takes his skill with him and inherits the car rating of wherever you put him, which is the only reason this page can answer a rumour.
The 2027 car ratings are the 2026 constructors’ order compressed toward the mean, because year two of a rule set is historically when a field closes up. The teams table shows the working for every one.
It changes the spread, not the order. At the default the grid runs from 91 down to 66, a spread of 25 against 2026’s 35. Pushing it up pulls every car toward the field mean, so the ratings keep their ranking but the gaps close; pulling it down widens them again. It is a statement about the season, so it moves every car deterministically rather than drawing anything.
A driver with no F1 mileage does not get a lower rating, he gets a wider one. Each driver carries a band around his skill prior, and a rookie’s is multiplied so that both a very good season and a very bad one stay live. That is why promoting an F2 front runner widens the whole distribution instead of simply lowering it, and why the lever does nothing until one is actually in a car.
Ten thousand seasons, each one twenty four rounds plus ten sprints, scored on the real points table. Every draw comes from a seeded generator, so the same grid and the same seed produce the same answer on any machine. Form is not redrawn from scratch each round: skill and car rating both take a mean reverting walk across the year, so a driver has runs of form rather than twenty four independent coin flips.
Injury, illness and bans are in only as a blunt per driver chance of missing rounds. Stewards are not modelled at all: no penalties, no disqualifications, no appeals, no exclusions for a floor that failed scrutineering. Nor is anything off track, from a team losing a title sponsor to a protest that rewrites a result three weeks later. The model prices a car, a driver and a season. It does not price the FIA, and any number here should be read with that missing.
Each one is a grid edit, not a setting. Flip it and drivers actually move, which is why it stays in step with anything you change by hand.
The round after which the title is mathematically over. A clinch needs the leader’s gap to exceed every point still on the table.
Matched seeds. Each remaining round is run three ways from identical draws: as is, the leader forced to win, the leader forced to retire. The bar is how far that moves the clinch.
assets/ folder.