The start, scored

A top-tier start here is a final score of 6.2 or better: the score a start earns, against the fleet reference, when it sits at the front third’s median on every measurement. Scores are read against the J/70 class reference (9227 starts from 52 sources). Sagara produced a top-tier start in 5 of 8 races, the most in the J/70; the middle boat of the fleet in 25% of its races; OceanPact in 5 of 8. The stricter reading, a fixed 7, was reached in 13% of the class’s starts, most often by Midlife Crisis (3 of 7).

The top 10% of finishers started with a median score of 6.5, the rest of the fleet 5.1; 58% of the top finishers’ starts were top tier, against 25%. Within races, the rank correlation between final start score and finishing place was -0.43 (-1 would be every race finishing in start-score order). Against the place two minutes after the gun it was -0.69, but the final score includes the ground made up the wind by + 2 minutes, so that association is partly mechanical; the score at the gun, from line and speed alone, gives -0.52. One point of the score at the gun (the provisional score: line and speed, known before the height is) was worth 4.2 ± 0.8 places gained at + 2 minutes and 2.7 ± 1.0 at the finish (379 starts in 8 races; ± is 1.96 standard errors, which ignores that a boat’s starts are alike from race to race; 12 starts over at the gun left out).

A boat that held its lane and was in clear air at + 60 s was in the top third two minutes in 48% of the time (74 of 153); one that was rolled, 25% (1 of 4); one that tacked away, 19% (19 of 100); one that bore away, 2% (1 of 43). These are associations across one event’s starts, not a measured effect of the start alone: a boat that starts well is often a boat that is fast.

The score

Each start is measured in the start line’s frame from the boat’s tracked positions, resampled to one second, with the committee-marked line ends at the gun and the race’s wind axis (the course the fleet sailed, not a measured wind). Seven measurements are read as percentiles of a reference fleet and turned so that higher is better; they make three groups:

• Crossed after the gun (line)
• Behind the line at the gun (line; 0 when over)
• Speed at the gun (speed)
• Speed held into + 10 s (speed)
• Ground up the wind by + 2 min (height)
• Angle to the wind, + 10 to + 60 s (height)
• Line bias given up (height)

Line weighs 40%, speed 30% and height 30%; the score is 1 + 9 × the weighted mean of the groups, to one decimal. A boat over the line at the gun, by the tracker or by an official OCS, UFD or BFD, scores 1. The provisional score uses line and speed only, as a boat’s device would show it at the gun; the final score adds height at + 2 minutes. The definitions are the start score’s specification (start-score-1), the same that runs on the boats’ phones and watches.

The reference

The class reference, built from measured championship fleets, is used when it holds at least 300 starts (a front-third reference at least 100); otherwise the class is scored against this event’s own starts (front third by finishing place), leaving out races in which more than 40% of the fleet was over at the gun. For the J/70 the fleet reference is the J/70 class reference (9227 starts from 52 sources), and the front third’s is the J/70 class reference (2886 starts from 52 sources). Fleet: 9227 starts from 52 sources. Front third: 2886 front-third starts from 52 sources. A score against an event reference says how a start compared with the fleet on the day.

The data

2026 J/70 North American Championship, Chesapeake Bay, Annapolis, Maryland · Annapolis Yacht Club, 29 October 2026 – 31 October 2026: 8 J/70 races measured, 396 starts. Positions are the organiser’s trackers at about one fix a second; the start-line ends are the committee’s marked positions, recorded every 3 s, and the gun is the published start time. The committee end is the line’s starboard end looking up the course. Height is scored only where the first leg is a beat.

Lane, shadow and place

• Snapshots at + 10, + 30 and + 60 s. A neighbour is a boat within 10 hull lengths along the wind; the gap to windward and to leeward is measured across the wind, windward being the side the wind comes from on the boat’s tack.
• The shadow is the snapshot wake model used in our start reports, run on each boat’s tracked velocity in a wind of 12 kt on the course axis; a deficit of at least 1% is bad air. The same is run at 8, 12, 16 kt and -10, 0, 10° off the axis, and the table shows the deciding case. It is modelled only when at least 90% of the starters have a position at the time.
• Held lane: from the first moment the boat has held a close-hauled course on one tack for 5 known seconds, no tack and no bear-away before + 60 s; a boat that never settles close-hauled bore away at the gun. A tack is the other tack held for 5 s; a bear-away is a course more than 15° below the race’s close-hauled angle (the median of the starters’ median angles from + 10 to + 60 s) held for 5 s. Seconds below 2 kt carry no course; a boat with fewer than 40 known seconds is unknown.
• Outcome, the first that applies: rolled; tacked; bore away; held its lane in bad air at + 60 s; held its lane and in clear air at + 60 s. The air is read at + 60 s only: a boat classed clear may have been in modelled shadow at + 10 or + 30 s.
• Place at + t is the rank of the distance made along the course among the boats tracked then; metres behind are to the leader. Top third at + 2 minutes is a rank within the first third of the boats ranked then.
• Top 10% of finishers: a finishing rank within the first tenth of the boats ranked in that race. The value of a point is the slope of place on the provisional score (line and speed, known at the gun) within races, each race’s means removed; the final score is not used for it because its height is the ground made by + 2 minutes.

Caveats

Everything is measured at the tracker, not the bow: a boat counted over by half a metre may not have been called. The wind is the course axis, so height is relative to the course the fleet sailed. Speeds and angles are over the ground and include current. The shadow model is a prior on where the air was disturbed, not a measured loss of speed, and the ground velocities it uses are not the boats’ apparent wind. Finishing places are the official race places where they are published. Correlations describe this event and do not separate start quality from boat speed.