Lake Huron · Race committee brief
Which forecast should score the Bayview Mackinac Race?
Keep the published method for 2027 trophies. For trial scores, start with NOAA’s HRRR. Its archived forecast issues let the committee reproduce what it could have known at the gun. Predeclare the cycle and longer-range forecast for boats that outlast HRRR’s 48-hour extended runs. Run ECMWF and RRFS alongside it. The Lake Huron comparisons have yet to establish a consistent winner.
On July 10, 2027, the fleet starts one 225 nm course through a virtual gate south of Great Duck Island, finishing off Mission Point. The historical comparison keeps the 204 nm Shore and 259 nm Cove Island courses as sailed. Chair’s announcement · 2025 sailing instructions
What the routed scores changed
Pre-start HRRR routing changed 1 of 4 winners and 27 of 51 places in the 2023 Shore test. These four classes had a matched race-year ORC polar and a completed route for every official finisher. Lake Effect moved from fourth to first in Class M, ahead of Dos Mas. Official 2023 class results.
Four-class pre-start test
| Class | Official winner | Routed winner | Places changed |
|---|---|---|---|
| Class G | Snipe | Snipe | 5 / 13 |
| Class J | Courage | Courage | 8 / 12 |
| Class M | Dos Mas | Lake Effect | 11 / 13 |
| Class N | Pirate | Pirate | 3 / 13 |
All 49 hourly files of the July 15 12 UTC forecast were available by 14:03 UTC, before the starts. All modeled finishes were within its 48-hour horizon. In Class M, HRRR predicted 37.71 hours for Lake Effect and 36.50 for Dos Mas. The model used wind and each boat’s race-year polar. That 1.21-hour difference in predicted passage time is small relative to a passage of about 37 hours. Both hindsight wind analyses put Stalwart first on the restricted route library. Wind inputs and routing methods both changed, so this example cannot isolate their effects.
We matched race-year ORC polars by boat name and sail number. The sail configuration accepted by the committee could differ. For the separate hindsight test we replayed the wind on one or two routes checked against the marks and land mask. Each boat took its fastest candidate. We ranked official elapsed time divided by predicted time.
HRRR analysis
ERA5 reanalysis
HRRR put Diablo first and ERA5 kept Blitzkrieg. This class’s official factor spread was 3.54%, compared with 1.05% under HRRR and 1.90% under ERA5. The narrower allowances give elapsed time more weight. Across 69 placed boats in 7 complete class-editions, the two wind analyses picked different winners in 2.
Class winners and example boat ratings
| Year | Course / class | Official | HRRR | ERA5 |
|---|---|---|---|---|
| 2022 | Cove Island · Class C | Notorious | Carrera | Carrera |
| 2022 | Cove Island · Class D | Blitzkrieg | Diablo | Blitzkrieg |
| 2022 | Cove Island · Class F | Fast Tango | Smokum Too | Smokum Too |
| 2023 | Shore · Class G | Snipe | Snipe | Snipe |
| 2023 | Shore · Class M | Dos Mas | Stalwart | Stalwart |
| 2023 | Shore · Class N | Pirate | Pirate | Knot Yours too |
| 2026 | Cove Island · Class C | Eagle One | mc 2 | mc 2 |
We omitted 6 of the original 13 complete classes. We excluded the 2021 and 2022 Shore hindcasts from both models. Each year had only one accepted track and the median observed passage took about 71% of the HRRR prediction. For the ten J/35s in 2021 Shore H, the effective factors (corrected divided by elapsed time) also differed by less than 0.0001, consistent with level scoring. The retained comparisons cover 2022, 2023 and 2026. Four classes had two accepted paths. Switching every boat from one path to the other changed the leader in six of eight class-by-model comparisons.
2022 Cove Island Class D
Modeled factors divide mean predicted class time by each boat’s predicted time, so compare relative spreads within each column.
| Boat | Elapsed, h | Official factor / place | HRRR factor / place | ERA5 factor / place |
|---|---|---|---|---|
| Blitzkrieg | 52.55 | 1.0142 / 1 | 0.9958 / 2 | 0.9898 / 1 |
| Diablo | 52.22 | 1.0327 / 2 | 0.9956 / 1 | 1.0012 / 2 |
| Unplugged | 53.24 | 1.0185 / 3 | 0.9982 / 5 | 0.9922 / 4 |
| Capers | 52.59 | 1.0366 / 4 | 1.0005 / 4 | 1.0014 / 3 |
| Variance | 52.52 | 1.0479 / 5 | 1.0010 / 3 | 1.0058 / 5 |
| Pterodactyl | 53.52 | 1.0359 / 6 | 1.0006 / 6 | 1.0027 / 6 |
| Freedom | 54.37 | 1.0340 / 7 | 1.0010 / 7 | 1.0019 / 7 |
| Relentless | 56.69 | 1.0121 / 8 | 1.0013 / 9 | 0.9966 / 8 |
| Pendragon | 56.20 | 1.0474 / 9 | 1.0060 / 8 | 1.0087 / 9 |
Before the 2026 start, the available 12 UTC HRRR and 06 UTC Euro IFS runs reversed 2 of 35 within-class pairs across 17 matched boats. Among seven of nine boats in Shore Class L, HRRR put Roxy first and IFS put My Way first. All four comparison classes had missing boats. The winds were issued before racing and the candidate paths were selected in hindsight.
How close were the class wins?
In 2025 Bayview used a fixed course wind mix to set each boat’s time-on-time factor. Multiplying elapsed time by that factor gives corrected time. Across 2,023 finishes in 185 eligible classes, a 1% increase to the winner’s factor alone would have put the runner-up ahead in 74 classes. Bayview scoring rule
Shore
Class leads reversed out of 110
Cove Island
Class leads reversed out of 75
In 2025, Wind Toy IV beat Shape by 3 min 11 s in Shore Class H. Increasing Wind Toy IV’s factor by 0.23% reverses the pair. On Cove Island, Heartbreaker’s Class A margin over Mockingbird was 74 seconds. Official 2025 results.
Results by course and year
| Year | Course | Finishers | Classes | Leads reversed at +1% |
|---|---|---|---|---|
| 2011 | Cove Island | 39 | 5 | 1 |
| 2011 | Shore | 127 | 10 | 4 |
| 2012 | Cove Island | 33 | 3 | 0 |
| 2012 | Shore | 110 | 9 | 5 |
| 2013 | Cove Island | 31 | 4 | 2 |
| 2013 | Shore | 116 | 10 | 3 |
| 2014 | Cove Island | 24 | 2 | 1 |
| 2014 | Shore | 121 | 9 | 2 |
| 2015 | Cove Island | 97 | 8 | 2 |
| 2015 | Shore | 57 | 4 | 3 |
| 2016 | Cove Island | 80 | 6 | 2 |
| 2016 | Shore | 66 | 5 | 1 |
| 2017 | Cove Island | 79 | 6 | 5 |
| 2017 | Shore | 62 | 5 | 3 |
| 2018 | Cove Island | 84 | 8 | 5 |
| 2018 | Shore | 52 | 6 | 2 |
| 2019 | Cove Island | 78 | 9 | 5 |
| 2019 | Shore | 48 | 5 | 2 |
| 2020 | Shore | 79 | 11 | 4 |
| 2021 | Shore | 104 | 10 | 2 |
| 2022 | Cove Island | 59 | 7 | 4 |
| 2022 | Shore | 65 | 6 | 1 |
| 2023 | Cove Island | 55 | 4 | 1 |
| 2023 | Shore | 96 | 8 | 4 |
| 2025 | Cove Island | 82 | 6 | 2 |
| 2025 | Shore | 80 | 6 | 3 |
| 2026 | Cove Island | 53 | 7 | 3 |
| 2026 | Shore | 46 | 6 | 2 |
28 course-editions from 2011–26. Scoring systems varied over these years. We used the organizer’s timed, ranked results in handicap classes with distinct effective factors and a recorded first and second place. Multihulls and the 2024 Centennial course are outside this calculation. Bayview results archive.
Which weather should set the factor?
ORC used weather routing in the Tallinn Worlds offshore race. Transpac’s 2025 forecast factors used Expedition with HRRR and GFS. The rating rewards performance against each boat’s predicted passage time. ORC Weather Routing Scoring · Tallinn Worlds offshore example · Transpac’s 2025 implementation
HRRR vs Euro before the gun
2025 start-day 00 UTC forecasts, 24–48 hour leads. Wind-vector error in m/s, lower is better.
| Lake Huron | HRRR | Euro IFS |
|---|---|---|
| North · 45003 | 3.28 | 2.52 |
| South · 45008 | 3.17 | 3.27 |
| Alpena · 45162 | 3.55 | 2.99 |
75 paired station-hours. The better model depended on where the fleet sailed.
Did the Euro ensemble help?
2025 pre-start forecast, three buoys at 24, 36 and 48 hours. Wind-vector error in m/s.
Averaging the winds reduced vector error by 25% in 9 buoy comparisons. For scoring, we routed the control and four perturbations separately in 2026. The control matched the official order in 54 of 61 boat pairs. Averaging predicted passage times matched 52. This measures agreement with published results. Use member spread to flag uncertain awards. ECMWF archive.
RRFS deserves a trial. For the 2023 start, a later NOAA rerun scored 4.17 m/s against same-cycle HRRR’s 5.37 across 90 paired buoy-hours. The rerun was published in 2025. One race window cannot establish which model should set official factors. NOAA RRFS archive.
Forecast coverage and wind checks
Ensemble passage times. Across 28 matched boats in 2026, the control’s mean absolute passage-time error was 2.00 hours and the five-member mean’s was 2.07. Only Cove Class C had every finisher matched. Four members put Eagle One first and one put mc 2 first. The mean retained Eagle One, the official winner. Partially matched fleets can answer pair comparisons but cannot establish whole-class winners.
Other ensemble checks. Across four starts and 55 buoy comparisons, the NOAA GEFS mean reduced vector error from 3.51 to 3.21 m/s against its control. NOAA GEFS.
HRRR and Euro. Specify the model cycle, publication cutoff and longer-range fallback before racing. In July 2026 at Alpena, same-cycle Euro IFS scored 2.00 m/s and HRRR 2.21 over 15 readings. A newer HRRR run scored 1.98. Issued forecasts.
RRFS. Operational RRFS began after the July 2026 race. In a September prototype test at buoy 45154, its height-adjusted vector error was 1.88 m/s against HRRR’s 2.56. Only 18 of 54 planned observations matched, all at one station. NOAA archive.
ERA5. Reanalysis uses later observations. Across 749 race-window buoy hours in six editions, hindsight HRRR and ERA5 were close overall (2.54 and 2.58 m/s). HRRR led at northern buoy 45003 and ERA5 at southern buoy 45008. Wind checks adjust buoy winds to 10 m. Buoy archive · ERA5 archive.