In short. Our estimated finishes divide the distance each boat has left by its speed towards the finish over the last 24 hours, then add the lap of Guadeloupe. We replayed that calculation over the race's 155 position reports, from 1 to 6 November. It is usually off by 14 to 18% of the time left, more often late than early. Our first range was too narrow: it held the actual time only once in four. It now runs from 20% earlier to 45% later, and it is right at least three times out of four.
A division, not a crystal ball
At every report, every hour, the official tracker publishes how far each boat is from the finish. It also publishes its VMG, the speed at which it is actually closing on the finish. That is not boat speed: a trimaran doing 30 knots at 45° to its course only gains 21 knots towards Pointe-à-Pitre.
The estimated time is that distance divided by the VMG. The question is which VMG:
- The VMG over the last 24 hours counts for 70%. It smooths out tacks, gybes and the lull of a windless night.
- The VMG over the last 4 hours counts for 30%. It tells whether the boat has just sped up on reaching the trade wind, or slowed under a ridge.
We tried other mixes on the same data. The 24-hour VMG alone does almost as well. The 4-hour VMG alone does clearly worse beyond a day, because it mistakes a squall for a trend.
The site gives no estimate for a boat not heading for the finish (VMG under 2 knots), retired or reporting damage. The time is rounded to the quarter hour and recalculated at every report.
The last lap of the island counts separately
The finish is not judged in open water. After Tête à l'Anglais, north-west of Basse-Terre, boats run down the lee coast, round the Basse-Terre buoy and then sail back up to Pointe-à-Pitre. That lap is about 56 miles. The mountains steal the wind there, and nobody sails it at open-ocean pace (see the weather traps from Saint-Malo to Guadeloupe).
So the estimate counts those last 56 miles at a speed set for each class: 13 knots for Ultims, 10 for Ocean Fiftys, 9.5 for IMOCAs, 8 for Class40s, 7 for Vintage Multis and 6 for Vintage Monos. These are our assumptions, drawn from past editions, not 2026 measurements. The first finishes will tell whether they hold.
How we measured accuracy
No boat has crossed the line yet. To judge the method, we took each actual position of a boat, 6 to 72 hours after a report, as a dummy finish line. At the starting report, the site's calculation says how long the boat will take to cover that distance. We then compare with the time it really took.
The whole fleet goes through it, over the 155 reports from 1 November at 12:02 to 6 November at 12:00 UTC: more than 30,000 comparisons. Only the first day is left out, since the 24-hour VMG does not exist yet.
Actual time against estimated time
- 6 h 77%before 45%
- 12 h 77%before 26%
- 24 h 79%before 25%
- 48 h 83%before 33%
- 72 h 86%before 35%
See the figures
| Horizon | Cases | Median | Half of cases | Inside the range |
|---|---|---|---|---|
| 6 h | 9,040 | +2% | −10% / +22% | 77% (before 45%) |
| 12 h | 8,428 | +4% | −9% / +22% | 77% (before 26%) |
| 24 h | 7,199 | +6% | −7% / +23% | 79% (before 25%) |
| 48 h | 4,786 | +11% | −1% / +27% | 83% (before 33%) |
| 72 h | 2,379 | +17% | +5% / +31% | 86% (before 35%) |
Three lessons:
- The typical error does not shrink with distance. It is 14% of the time left at 6 hours from the dummy line, 15% at 24 hours and 18% at 72 hours. An estimate two days out is therefore commonly off by seven hours.
- The error leans late. In the median case, the boat takes 2% longer than predicted at 6 hours, and 17% longer at 72 hours. Over this first week the fleet slowed on reaching the Azores high, and the previous days' VMG promised too much.
- The errors are lopsided. A boat rarely arrives more than 20% early, but a windless transition or a breakage can delay it by 50% or more.
What the measurement changed
The first version of the range widened by 6% plus 2.5% per day left, slightly more on the late side. It was a reasoned choice, not a measurement. Replayed, it held the actual time only 25 to 45% of the time.
Since this article, the range comes from the measurements: from 20% of the time left earlier to 45% later. It holds the actual time 77% of the time at 6 hours, 79% at 24 hours and 86% at 72 hours. It is wide, on purpose. A tight range that is wrong three times out of four gives false comfort.
In practice, 32 hours from the finish, an Ultim gets a range of about six and a half hours earlier to 15 hours later. A Class40 nine days from the line gets nearly two days earlier to four days later: at that distance, only the day of arrival means anything.
Accuracy varies a lot between classes. Here are the figures for a dummy line 24 hours away:
| Class | Typical error | Inside the range |
|---|---|---|
| Ultim | 28% | 60% |
| Ocean Fifty | 19% | 70% |
| IMOCA | 19% | 69% |
| Class40 | 13% | 86% |
| Vintage Mono | 13% | 85% |
| Vintage Multi | 14% | 83% |
Why Ultims are the hardest to forecast
An Ultim's speed depends enormously on the wind: it tops 30 knots in the trade wind and drops below 10 in a lull. The slightest change in the weather therefore upsets its VMG, while a Class40 keeps a steadier pace. Ultims also make the biggest route choices: going round a high to the west can make VMG collapse for half a day, until they find wind again.
Tom Laperche, leading on SVR-Lazartigue, shows it well. Here is what our estimate said, recalculated with the current range, at a few reports of the race (Guadeloupe time, five hours behind Paris and four behind UTC):
| Report | To go | Estimated finish and range |
|---|---|---|
| Mon 2, 8:00 | 2,832 miles | Fri 6, 5:45 Thu 5, 11:00 → Sun 8, 0:00 |
| Tue 3, 8:00 | 2,056 miles | Fri 6, 2:15 Thu 5, 13:00 → Sat 7, 8:15 |
| Wed 4, 8:00 | 1,723 miles | Sun 8, 4:15 Sat 7, 9:45 → Mon 9, 21:45 |
| Thu 5, 8:00 | 1,095 miles | Sat 7, 7:45 Fri 6, 22:15 → Sun 8, 5:15 |
| Thu 5, 20:00 | 712 miles | Fri 6, 23:30 Fri 6, 18:00 → Sat 7, 12:00 |
| Fri 6, 8:00 | 575 miles | Sat 7, 16:45 Sat 7, 10:15 → Sun 8, 7:30 |
Between 3 and 4 November, SVR-Lazartigue cut its distance to the finish by only 87 miles in twelve hours, against nearly 400 in the half-day before: the trimaran was going round a windless zone. The estimated time slipped by two days, then came back as the boat found the trade wind. The swings come from the weather, not the arithmetic, and only a wide range reports them honestly.
Recent speed or routing: two opinions on the same finish
The estimate described here assumes the boat keeps its recent speed to the end. For the top five of each class, the skipper card of the live tracker also gives a routed finish. It is computed by sailing the boat through the forecast wind with its speed polar. How it works, and its own accuracy, are explained in weather routing on the live tracker, how it works.
The two opinions often differ. On 6 November at 8:00 Guadeloupe time, the constant-speed estimate expected Tom Laperche on Saturday 7 around 16:45, and the routing on Sunday 8 around 3:30, some ten hours later. That gap means something: the wind forecast for the end of the course does not look like the last 24 hours. When the two agree, the time is safer. When they differ, the truth often lies in between, and the range usually covers both.
The limits, frankly
- One race, one weather pattern. Our measurements cover five days of the 2026 Route du Rhum. Boats in the same class sailed the same wind, so their errors are not independent. Another week, in a settled trade wind, the estimate would be more accurate.
- Dummy finishes. A position in mid-Atlantic is not the Pointe-à-Pitre line. The lap of the island has not been measured yet.
- Breakages cannot be forecast. A boat slowing down for repairs leaves the range, and stays out until its 24-hour VMG catches up.
- The data comes from the official tracker. If a report is missing or late, the estimate starts from an older position.
We will measure again with the real finishes, and recalibrate the range if needed.
Where to find the estimates
- The whole fleet at a glance: the estimated finishes page ranks boats by likely finish time, with their range and a countdown, in Paris time, Guadeloupe time or your phone's time.
- One particular boat: its card on the live tracker gives the estimate, the routed finish and a button to add the finish to your calendar.
- The class calendar: the record-based reference dates are in when do the boats arrive in Guadeloupe?.
Frequently asked questions
How is a boat's estimated finish time calculated?
The distance the boat has left is divided by its speed towards the finish (VMG), weighted 70% on the last 24 hours and 30% on the last 4. The last 56 miles, the lap of Guadeloupe, are counted at a slower speed set for each class. The calculation is redone at every report, every hour.
Is the estimated finish time reliable?
It is usually off by 14 to 18% of the time left, about seven hours two days out. The displayed range, from 20% earlier to 45% later, holds the actual time at least three times out of four according to our measurements from 1 to 6 November 2026.
Why is the range wider on the late side?
A boat rarely beats its recent speed by much, but a windless zone, a ridge or a breakage can slow it down sharply. On the 2026 Route du Rhum, the actual time ran over the estimate more often than it came in under.
How does it differ from the routed finish?
The estimated time extends the speed of the last hours. The routed finish sails the boat through the wind forecast by the GFS model. When the two agree, the time is safer; when they differ, the coming weather does not look like the last 24 hours.