How the Bermuda Race is Won

Stream strategy is not “go east” or “go west.” It is a time-dependent vector problem: forecast the current, find the meanders and rings, and cross the fast water without being carried away from the finish. The race’s own navigator resources make the same point: the Stream changes enough that a climatological route is not a race plan.

The tracks agree. The fleet’s longitude at 37°N shifts substantially from edition to edition, and the good boats move with the fleet rather than occupying one sacred meridian. A tracker shows the decision; without a contemporaneous current analysis it cannot tell us whether the decision was correct.

That accounting became part of the result in 2024. Newport Bermuda was the first ocean race to use forecast-time correction factors (F‑TCF): before the start, each boat was optimally routed through the expected wind and current, and its predicted elapsed time set its allowance. The rating asked a new question—how hard was this particular moving ocean for this particular boat?—without changing the navigator’s job aboard.

For every timestamped leg we subtract HYCOM’s surface-current displacement from the tracker’s ground displacement. The result is an estimated distance through the water; projecting the same current along the leg gives the ocean’s contribution. It is accounting, not a speed inference.

The 2024 field paid the median boat — nm, but route choice mattered: the middle 80% spanned — to — nm. In the controlled route family, bending modestly west before the crossing gained enough current to erase the detour and roughly — nm more besides.

More assistance did not reliably buy a better corrected place. After centering boats within year and class and controlling for distance sailed, the association was tiny (partial r = —, —, n = —). That is a useful warning, not a failed result: strong current is only valuable when it points somewhere useful and the windward cost of reaching it does not consume the dividend. WX routing must account for the whole bargain.

Within the same edition and published class, extra distance has a small but stubborn association with a worse corrected place: r = — across — boat-races; — after shuffling places only within class and year.

Small is the important word. Distance is not destiny: a mile sailed quickly in favorable wind or current can beat a shorter mile sailed slowly. But a routing excursion begins in debt. Across these races, unnecessary mileage did not disappear in the rating arithmetic.

Doublehanded racing has been part of every tracked edition here: 99 finishers, from 13 to 21 boats a race. The obvious story is fatigue—two sailors must steer, trim, navigate, eat, and sleep—so their tracks ought to wander farther.

They do not. Their median log is 0.4 nautical miles shorter than the edition median; the fully crewed boats sit at zero. The difference is negligible and statistically unpersuasive (—). That is not proof of equal performance: the fleets contain different boats and ratings. It is a narrower, nicer finding. Two people leave no obvious scar in the geometry of the track.

Being unusual was weakly costly, too. Across 3,096 boat-latitude observations, the farther a boat crossed from its own class median at 39°, 37°, or 35°N, the worse its within-class finish tended to be (r = —; — in a 10,000-draw permutation confined to class, year, and latitude). The effect is tiny. The useful result is not “follow the herd”; it is that a lonely route needs a real weather or current edge to repay its separation.

The cloud narrows as Bermuda rises over the bow. That convergence can make the race look decided late, but it is mostly geometry: everyone must arrive at the same island. The consequential separation happened farther north, while the boats still had room to trade distance against wind and current.

The practical lesson is modest. Measure every detour against what it buys now; distrust any gate remembered from the last race; and keep enough crew—whether two or twenty—to execute the new answer when the ocean changes the question.