Organizers of the Newport Bermuda Race say they will review whether to continue using forecast-based scoring, following an investigation into the error that scrambled results in June.
The Bermuda Race Organizing Committee and US Sailing released a 14-page report Tuesday tracing the error to two causes: conflicting weather files left active in routing software on race morning, and a separate defect in that software. The report attributes most of the impact to the first.
The committee said it will address three broader questions in a later review — whether forecast-based scoring improves fairness, whether its complexity outweighs its benefits, and whether the ORR rating rule is the right foundation for the race. The report does not say when that review will conclude.
What the report found
Forecast Time Correction Factors, or F-TCFs, are handicap numbers calculated by running each boat’s performance model through routing software called Expedition, using pre-selected wind and current forecasts. For this race, those were an HRRR wind file for the coastal start, a GFS file offshore and an RTOFS file modeling the Gulf Stream.
On the morning of Friday, June 19, the operator also downloaded ECMWF and Mercator files from within Expedition, intending to use them after the race to compare forecast models. Those files remained active during the official run, leaving the software working with two competing wind sources and two competing current sources.
Expedition developer Nick White described the result in the report. The software selects data based mainly on model run time and resolution, he said, which works when files complement each other, but if two comparable models are loaded from the same run time, “which gets used is effectively random.”
US Sailing’s standard practice is to close and restart Expedition immediately before an official run to clear extraneous data. That step “was apparently not performed,” the report says. A trial run the previous day using standard procedures produced accurate results.
The routes that came out looked plausible, and a sample check against boats’ certificate ratings did not raise concerns, partly because a ridge of high pressure near the course was already expected to move some boats’ numbers away from their ratings. The report calls the checks in place “not sufficiently robust.”
Four days before anyone flagged it
The error was not reported until Tuesday, June 23, when a competitor asked US Sailing to review two boats expected to perform similarly that had been assigned substantially different numbers. The competitor owned both — Oakcliff’s two Farr 40s.
Re-running the two boats individually produced markedly different results. Re-running the full fleet produced different results again. Race officials were notified that Friday’s numbers were wrong, and all 140-plus boats were re-run individually overnight. Corrected factors reached the Race Committee on Wednesday, June 24.
After several days of redress hearings, the International Jury directed publication of corrected results on June 26. Those have since been verified through multiple independent re-runs by different operators using separate instances of the software.
The software defect
The investigation also identified a genuine bug: Expedition produced different results routing boats in batches than one at a time. White confirmed it and released a corrected version.
Because that software had scored previous races, competitors asked whether earlier results were affected. BROC compared the two sources of error and found the defect alone produced deviations roughly one-fifth the size of the June 19 error, affecting far fewer boats. The report concludes it is unlikely to have materially changed the outcome of prior races.
The report also notes testimony at a redress hearing that a competitor was aware of an anomaly between the software’s two modes 30 to 60 days before the race, and said questions had been raised about it in an online forum of Expedition users. White reported that he first learned of the possible defect from US Sailing on June 23.
Recommended changes
The committee’s Technical Committee recommends nine corrective actions. They include requiring the software to be fully restarted before any official run with no other operations performed first; a written step-by-step procedure operators must follow and log; independent verification on two instances of the software by different operators before release; objective rather than visual quality checks; a time buffer between generating the numbers and the start; and retention of output logs from every official run.
The committee also wants Expedition’s developer to add a warning or lockout when competing forecast files are active for the same area and period.
The investigation was limited to the technical failure and does not examine actions by the Race Committee or the International Jury.
“BROC and its Technical Committee remain fully accountable for the results of the race and regret the disruption and uncertainty the F-TCF error caused competitors,” the report states.
The report closes out the 2026 organizing committee. Planning for the 2028 race passes to chair Patti Young and vice chair Will Passano.
The full report is available at bermudarace.com.
