MASSA 2017 Cressy Qualifier Radial

Performance rating asks, for each skipper: holding every other boat fixed, which rating makes my actual race-by-race finishes least surprising? Boats are valued with the singlehanded rating rather than the fleet rating: 0.79 × the skipper's fleet rating from the week before the event (unrated skippers start at 1.11); the skipper's own singlehanded offset. Strength differences are scaled by 0.80 before they become probabilities, the value that calibrates the model on seasons it had not seen. Non-finishes (OCS, DNF and similar) are left out of the fitting, as in the rating model.

Percentile places a rating among the 495 ISSA sailors with a singlehanded rating who raced the discipline in the last three seasons. Surprise is how many bits more surprising the finishes were at the pre-event rating than at the performance rating. P(this good or better) simulates the event 3,000 times at the pre-event ratings, with the same regatta-form model as the fleet analysis pages, and reports how often the skipper would score this many points or fewer in the races they sailed.

The singlehanded offsets are the current ones (the discipline files keep no weekly history), so a skipper's own offset has already seen this event; offsets are shrunk hard, so the effect is a few hundredths at most. Only skippers are rated.

• ^u no fleet rating — the model's level for unrated skippers is the baseline.
• — in the performance column: last in every race sailed, so no finite least-surprisal rating exists.