How good are these forecasts?
Every forecast scored on seasons the model never saw while its settings were chosen.
Each race is ranked on its own: a team's rating is its sailing programme (skipper, tactician, crew) plus a boat term for the season, since new TP52s are built most years.
The series scores low points: a team's race place is its score, a regatta adds up its races (coastal races can count more) and the season adds up the regattas. The title odds sail the rest of the season race by race and add the places the same way.
Wind is the weather that matters here. Light air makes a race more of a lottery (shifts and holes decide more than boat speed), and so does a strong breeze; the model tests how much either should flatten the ratings.
Penalty points and redress awards are left out of the ratings (a redress race has no real place); the season table uses the series' own totals where they are published.
Race dates within a regatta come from the published schedule, so a race is placed on the day it was most likely sailed.
Questions and answers
How accurate are PodiumOracle's 52 SUPER SERIES predictions?
On 125 races from held-back seasons (2024 to 2026), which the model never saw while it was tuned, its favourite won 16% of the time in fields of about 12, and it gave the eventual winner 10% on average. Its log loss was 2.437, against 2.503 for treating every team as equal and 2.452 for recent form.