Glossary

What the numbers on these pages mean.

Rating: A boat's overall strength: team, driver, recent form and penalty record added together. The fleet averages 1500, and 100 points is roughly a 64–36 edge in a head-to-head race.
Team, driver, form and penalty parts: The pieces a rating is built from, each shown as a distance from the fleet average; they add up to the rating. The driver part is what the model attributes to who is steering rather than the boat.
Skill (drivers): How much better or worse a driver makes a boat than an average SailGP driver would. It relies on drivers who have sailed for more than one team, so a driver who has only ever raced one boat has a less certain rating.
Recent form: Average finishing place over a team's last 30 races, normalised so 0 is always first and 1 is always last. On its own this was the single strongest predictor we tested.
Place in the fleet: A finishing place rescaled by fleet size: 0 for a win, 100 for last. It lets races with different numbers of boats be compared.
Chance of winning: How often a team wins in simulated racing. Event odds come from simulating the remaining races; season odds simulate every remaining event and the Grand Final.
Expected points: Average season points across all simulations, not a prediction of any single outcome.
Fleet points: Points from the fleet races only, before the final. These decide the season standings and who makes the final.
Deserved result: Where a team would have finished an event ranked on fleet points alone, ignoring the final. The gap between deserved and actual is how much the final changed things.
Wind class: Each race is labelled light, medium or strong from modelled wind at the venue, using the middle third of all SailGP racing as medium. Choppy is the equivalent for sea state.
Log loss: How badly a set of probabilities was wrong, in one number. Lower is better; a model that always said 'anyone could win' scores the uniform figure we compare against.
Calibration: Whether things predicted at 30% happen about 30% of the time. Being well calibrated matters more than being confident.
Elo: The scale ratings sit on, borrowed from chess: a fixed points gap always means the same win probability, whoever the teams are.