---
title: "Route du Rhum: how our weather routing finds the best route"
description: "GFS wind, fleet-calibrated polars, isochrones rerun every hour: how the live map draws the suggested route, its corridor and the routed arrival."
url: "https://www.offshoreraces.app/en/blog/route-du-rhum-weather-routing-how-it-works"
lang: "en"
---

# Route du Rhum: how our weather routing finds the best route

Published November 6, 2026 by [Antoine Labbe](https://www.offshoreraces.app/en/about)

![The IMOCA 11th Hour Racing in Lorient for the 2022 Défi Azimut](https://www.offshoreraces.app/photos/classes-imoca.jpg)

*Photo: Jug81, [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/), [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:11th_Hour_Racing_Imoca_%282021%29_au_D%C3%A9fi_Azimut_2022_02.jpg)*

**In short.** On the [Route du Rhum 2026 live map](https://www.offshoreraces.app/en/route-du-rhum/live), the **Routing** button shows the fastest route to Guadeloupe for the top five boats of each class. Every hour, a program takes each boat's latest position, the wind forecast by the American **GFS** model for the next five days and an approximate **polar** for its class, then searches for the quickest path using the **isochrone method**. The map draws it as a dashed route, a **corridor** in which the boat sailed three times out of four in our measurements, and a **routed arrival** that the skipper's card compares with the constant-speed estimate. It is an indicative routing, not the skippers' own: here is how it works, and how good it is.

## Why the route is not a straight line

Between Saint-Malo and Pointe-à-Pitre, the shortest route in miles is rarely the fastest. A sailing boat cannot sail straight into the wind, its speed depends on the angle at which it takes the wind, and the wind changes from place to place and hour to hour. Going around a windless area, such as the centre of the Azores high, adds miles but saves hours. That is the whole north-or-south question described in [the Route du Rhum weather traps](https://www.offshoreraces.app/en/blog/route-du-rhum-weather-traps-saint-malo-guadeloupe).

Routing answers a precise question: in the forecast wind, and with the speed this boat gets from each wind angle, which path brings it to the finish soonest? It takes three ingredients: a wind forecast, the boat's speed, and an algorithm to search for the route.

## The wind: five days of GFS forecast

The wind comes from **GFS** (Global Forecast System), the global model of NOAA, the US weather service. Its data is in the public domain. A new model "run" is published every six hours. We collect the wind 10 metres above the sea on a half-degree grid (about 30 miles) covering the North Atlantic, from 100° west to 20° east. The site keeps forecast steps every 3 hours up to 36 hours, then every 6 hours up to **120 hours**, five days. It is the same wind as the map's animated wind layer.

Between grid points the wind is interpolated in space, and between forecast steps, in time. The router can therefore ask for the wind at any position and any hour of the forecast.

## The boat's speed: one polar per class

A **polar** gives a boat's speed according to the wind strength and the angle between its course and the wind. The teams' real polars are secret. We use an approximate one per class: a top speed (36 knots for an Ultim, 24 for an IMOCA, 15.5 for a Class40), a ratio of boat speed to wind speed in light air, and a shape shared by each hull family. Multihulls are fastest with the wind on the beam, monohulls a little further downwind. Within 38° of the wind no boat moves forward: this is the **no-go zone**, which forces boats to tack to sail upwind. Above 35 knots of wind, speed drops, because the sea slows the boat more than the wind pushes it.

*Approximate polars of the Ultim and the IMOCA in 16 knots of wind blowing from the left: the distance from the centre gives the boat's speed for each wind angle. Below 38°, the no-go zone. Curves computed by the site's routing code, before calibration on the fleet.*

An approximate polar is bound to be wrong. At every run it is therefore **calibrated on the fleet**: for the first half of the class ranking, we compare each boat's real speed with the speed the polar gives in the wind it has on site. The median of these ratios becomes a factor applied to the whole polar, bounded between 0.5 and 1.35. At noon on 6 November, for instance, it was 0.89 for the Ultims and 0.82 for the IMOCAs: racing single-handed, boats sail below their theoretical potential.

## The algorithm: isochrones

To search for the route, the site uses the **isochrone method**, described by Japan's Hideki Hagiwara in the late 1980s and the basis of routing software ever since. From the boat's position, a series of headings is tried, every 5°, up to 110° either side of the direction of the finish. For each heading, the polar gives the speed in the wind at that time and place, and therefore the point reached after one hour (two for slower boats). From each of these points, the process starts again.

The number of points would quickly explode. At each step only the most advanced are kept: seen from the start, the horizon is cut into 1° sectors, and in each sector only the furthest point survives. The line of these points is the **isochrone**, the boundary of what the boat can reach in that many hours. Headings that cross land are discarded, using a coastline map at a tenth of a degree. As soon as a point can reach the finish, the path is traced back point by point to the start: that is the suggested route.

*The site's router in a 17-knot north-easterly trade wind, with a light-air zone on the straight line (dotted). From the start, the fan of headings tried; then one isochrone every 8 hours. They sag where the wind fails, and the fastest route (dashed) goes around the light air to arrive in 45 hours.*

The goal is not exactly the finish line: the computation stops at the northern approach to Guadeloupe. The lap around Basse-Terre to Pointe-à-Pitre is added at the route's average pace, slowed by a quarter for the island's wind shadows and manoeuvres. When the route does not reach the finish within the five days of forecast, the rest is extrapolated at the average rate at which the boat closed on the goal until then.

## Every hour, the top five of each class

The computation runs **every hour**, like the race position reports, from each boat's latest known position. It covers the **top five of each class** that are racing and located. For each class, that means between 4 and 8 million headings tried at each run. They fit in the allotted computing time: the top three are always routed, the fourth and fifth whenever the class budget allows, which has been the case so far.

The result is served together with the race positions: turning routing on costs the visitor's phone no computation.

## On the map: the route, its days and its corridor

The **Routing** button at the top of the map is off by default and remembers your choice. When on, it shows one route at a time, to stay readable: the selected boat's if it is among those routed, otherwise the leader's of the class you are viewing. In the "All" view, only the leader of each class is shown, as a thin line.

The route is a **dashed** line, so it cannot be mistaken for the boat's track, which is solid. One **dot per day** gives the forecast position and its time. Beyond the five days of forecast, the route becomes a **dotted** line to the approach to Guadeloupe. It then follows the course, round the north and down the west coast of Basse-Terre, to the finish line at Pointe-à-Pitre, where a flag marks the **routed arrival**.

Routed arrival ≈ Thu, Nov 12 · 01:38 AM

*Corentin Horeau's suggested route (IMOCA), computed at noon on 6 November 2026 from the 06 UTC GFS run: one dot per day, the corridor in which the boat sails three times out of four, the dots beyond the five days of forecast and, inset, the lap of Basse-Terre to the finish line. Paris time.*

Around the route lies a **corridor**. It is not an arbitrary margin: every computed route is stored, then compared hour after hour with the boat's real position. The corridor's radius is the one within which the boat sailed **three times out of four**, expressed as a share of the distance already covered along the route. It starts from the gap measured at 3 hours, widens with time, then narrows towards the finish, where all routes meet.

The routing legend, to the right of the map on a computer and in a sheet that slides up on a phone, sums it up: the routed arrival time, its gap with the constant-speed estimate, and the **measured accuracy**, the median gap between our routes and the boats' real positions at 6, 12 and 24 hours.

## On the skipper's card: the routed arrival

On a skipper's card, the estimated arrival time assumes the boat keeps its average speed to the line, with a range. If the boat is among the five routed in its class, a **Routing** line appears below: the routed arrival and its gap with the estimate, for example "+5 h" when the forecast wind will slow the boat. A marker places it on the range bar. The two figures answer two different questions: "if it carries on like this" and "if the wind does what the model forecasts". The whole fleet's arrival times are gathered on the [Estimated arrivals](https://www.offshoreraces.app/en/route-du-rhum/arrivals) page. How the constant-speed estimate is computed, and how far to trust it, is explained in [how we estimate finish times](https://www.offshoreraces.app/en/blog/route-du-rhum-estimated-finish-time-accuracy).

## How good this routing is

We measured its accuracy before showing it. Replaying the routes computed every 3 hours on 5 November for the top three of each class, the median gap between the routed position and the real one was about **20 miles at 6 hours, 40 miles at 12 hours and 67 miles at 24 hours**, 10 to 15% of the distance sailed. That is 2.5 times better than plain dead reckoning, which extends the course and speed of the last six hours.

The gap does not mainly come from the wind: the GFS forecast closely matches the wind measured on site. Two thirds of it comes from speed, which the approximate polar overestimates on a reach and underestimates in light air, and one third from the skippers' routing choices: they have other models, their real polars and what they see on the water. That is why the map shows a corridor rather than a single line, and why this routing remains **indicative**. Until the live tracker has enough measurements, the legend shows the figures from this test, then those measured during the race over the last four days.

To follow the routes for real, open [the live map](https://www.offshoreraces.app/en/route-du-rhum/live) and turn on **Routing**.

## Frequently asked questions

### What is weather routing in offshore racing?

Weather routing is the computation of the fastest route for a boat, given the wind forecast and the speed the boat reaches for each wind angle and strength (its polar). It is done by software that most often uses the isochrone method.

### Where does the wind used by Offshore Races' routing come from?

From NOAA's GFS model, in the public domain: wind at 10 metres on a half-degree grid, up to five days ahead, updated every six hours.

### How often is the route recomputed?

Every hour, from each boat's latest position, for the top five of each class in the Route du Rhum 2026.

### Does the routing show the route the skipper will follow?

No. Skippers have their own polars, other weather models and what they see on the water. Our route is the one a router would find with approximate polars and GFS wind. The corridor shows the measured gap: the boat sails within it three times out of four.

### Why does the routed arrival differ from the estimated arrival?

The estimated arrival assumes the boat keeps its average speed to the line. The routed arrival takes into account the wind forecast along the route: it moves earlier when the wind builds and later when it eases.

In this article

1. Why the route is not a straight line
2. The wind: five days of GFS forecast
3. The boat's speed: one polar per class
4. The algorithm: isochrones
5. Every hour, the top five of each class
6. On the map: the route, its days and its corridor
7. On the skipper's card: the routed arrival
8. How good this routing is
9. Frequently asked questions

## Keep exploring

Read next

- [Route du Rhum 2026: the list of retirements and their causesNovember 4, 2026](https://www.offshoreraces.app/en/blog/route-du-rhum-2026-retirements-list-causes)
- [Route du Rhum 2026: when do the boats arrive in Guadeloupe?November 4, 2026](https://www.offshoreraces.app/en/blog/route-du-rhum-2026-finish-guadeloupe-arrival-dates)
- [Route du Rhum 2026: the favourites, class by classOctober 30, 2026](https://www.offshoreraces.app/en/blog/route-du-rhum-2026-favourites-class-by-class)

Races in this article

- [Route du RhumRacing since Nov 1](https://www.offshoreraces.app/en/route-du-rhum)
  [Live](https://www.offshoreraces.app/en/route-du-rhum/live)[Standings 2026](https://www.offshoreraces.app/en/route-du-rhum/standings)[Estimated finishes](https://www.offshoreraces.app/en/route-du-rhum/arrivals)

## Races in this article

- **[Route du Rhum](https://www.offshoreraces.app/en/route-du-rhum)**
  The solo transatlantic race from Saint-Malo to Guadeloupe.
  Start: Saint-Malo · Finish: Pointe-à-Pitre · 3,542 nm · Racing since November 1, 2026
  [Ultim](https://www.offshoreraces.app/en/classes/ultim) [IMOCA](https://www.offshoreraces.app/en/classes/imoca) [Ocean Fifty](https://www.offshoreraces.app/en/classes/ocean-fifty) [Class40](https://www.offshoreraces.app/en/classes/class40)