Review · Rolex Middle Sea Race 2026 · offshore raceWe called a dead heat. The fleet found the coast.
Our Friday briefing said three routes for a First 40 would finish within an hour of each other. In the weather that actually happened, they finished within 13 min. Where the fleet beat our routes was close to land, on the first leg and at the Aeolian Islands, in detail our forecast was too coarse to see.
Published Covers our briefing of Friday 16 October, against the race
This review scores the briefing we published before the start, which stays online unchanged. Read the original briefing →
- What we said
- Three routes, effectively level, at about 110.4 h for a First 40 at 92% of polar, 8 in 10 forecasts between 105–121 hours. The one real choice was east or west on the leg from Pantelleria to Lampedusa, with east favoured.
- What happened
- In the weather that actually happened, our three routes took 106.6 h to 106.9 h, and the best possible route 104.8 h. The 22 boats rated like a First 40 took a median 106.4 h. East at Lampedusa paid, as we said.
- Where we missed
- Close to land. On leg 1 nearly the whole fleet sailed inside our routes, and boats further out did worse in 5 of 6 classes. At the Aeolians, 39 boats went north of 39°N, but none of our forecasts did.
What we published
This review evaluates the briefing as it stood before the start, which was our update of Friday 16 October, from the ECMWF ensemble forecast issued at 00:00 UTC that day then re-run for the official 11:10 start for IRC 5. Nothing we added on race morning, and nothing we re-ran once the boats were racing counts.
We routed a First 40 at 92% of its polar through all 50 versions of the forecast. Three routes stood out. The most likely went east of the rhumb line on the leg from Pantelleria to Lampedusa (fastest in 25 forecasts), the conservative went west there (17), and an alternative stayed inside on the first leg to the Messina Strait (8). We said it was too close to call and the typical finishes were within an hour of each other.
- Most likely: east at Lampedusa
- Conservative: west at Lampedusa
- Alternative: inside on leg 1
- Best in hindsight (ECMWF analysis)
- Best in hindsight (ICON-EU, sharper)
Did the decision hold up?
To judge the decision on its own, we ran our three routes again, with the same boat and settings, through the weather that actually happened each day. We used two pictures of it, the ECMWF analyses, stitched from the first hours of each new run, and the sharper ICON-EU model from Germany’s weather service which resolves the coast in much finer detail.
Our three routes, run again in the weather that happened each day
| Route | Our forecast, typical | Actual weather (ECMWF analysis) | Actual weather (ICON-EU, sharper near land) |
|---|---|---|---|
| Most likely east at Lampedusa | 110.4 h | 106.6 h | 108.3 h |
| Conservative west at Lampedusa | 110.4 h | 106.9 h | 108.0 h |
| Alternative inside on leg 1 | 111.3 h | 106.7 h | 106.2 h |
| Best possible, in hindsight | – | 104.8 h | 102.5 h |
First 40 at 92% of polar, currents included, from the 11:10 start. “Best possible” routes the same boat through the actual weather with perfect knowledge of it.
In ECMWF’s picture, the three routes finished within 13 min of each other, a dead heat, as we said. The best possible route, with perfect knowledge of the weather, was 1 h 50 quicker. Our typical time of 110.4 h was a few hours slow. The 22 boats rated like a First 40 took a median of 106.4 h, but spread from 96 to 129 hours, 10 were quicker than our 8-in-10 range and 2 slower. The weather explains the middle of that spread but as ever crew experience and perhaps some luck explain the rest.
ICON’s sharper picture tells a different story. In it, the alternative that stayed inside on leg 1 finished in 106.2 h, 2 h 04 ahead of the most likely route. The best route in ICON’s weather took 102.5 h. Near the coast there were a couple of hours to find, and our forecast couldn’t see them.
Leg by leg
We drew four lines across the course where the routes divide, and counted where the best route in each of our 50 forecasts crossed them, and where 89 IRC boats actually did.
Leg by leg: our 50 forecasts against the fleet
| Where | Our forecasts’ best routes (8 in 10) | The fleet (8 in 10 of 89 IRC boats) | What paid, within each class |
|---|---|---|---|
| Leg 1 off Siracusa37.0°N | 15.43°E–15.83°E | 15.35°E–15.46°E | Further out did worse in 5 of 6 classes |
| Messina Strait exit38.45°N | 15.40°E–15.53°E37 west, 13 east | 15.53°E–15.68°E0 west, 89 east | No clear pattern |
| Aeolians14.3°E | 38.73°N–38.90°N50 south of 39°N, 0 39°N or north | 38.68°N–39.11°N50 south of 39°N, 39 39°N or north | 39°N or north did better (median 31 in class), south of 39°N did worse (median 61 in class) |
| Lampedusa leg36.0°N | 11.91°E–12.38°E8 west, 42 east | 11.94°E–12.30°E13 west, 69 east | east did better (median 45 in class), west did worse (median 76 in class) |
“In class”: where a boat finished in its IRC class on corrected time, from 0 (class winner) to 100 (last finisher). Boats on one side may also have arrived at a different time, so this shows association, not cause.
Leg 1, off Siracusa. The fleet’s median crossing was 15.37°E, and the best routes in our forecasts ran from 15.43°E to 15.83°E. That’s nearly the whole fleet sailing inside our forecast, about 11 miles closer to Sicily than our typical route, and the inside track paid off within each class, boats further out did worse in 5 of 6 classes. The best route in ECMWF’s weather crossed at 15.47°E. In ICON’s sharper weather it crossed at 15.37°E, almost exactly where the fleet went.
Out of the Messina Strait, the best route in 37 of our forecasts turned west towards Stromboli sooner, but all 89 boats stayed further east. Results show no consistent pattern either way, so we score this one as neither right nor wrong.
At the Aeolian Islands, our routes all passed south of 39°N (north to 38.90°N). 39 of 89 boats went north of it, and they did better, with a median 31 in their class, against 61 for boats further south (on a scale from 0 for the class winner to 100 for last). Neither model’s best route in hindsight went north, so this was local knowledge, or local wind too fine even for ICON’s 4 NM grid.
On the leg from Pantelleria to Lampedusa, our routes split 8 west, 42 east, and the fleet split 13 west, 69 east, almost the same share. East paid off with a median 45 in class against 76 for the boats that went west. Several of the western boats arrived later, after a wind shift on Tuesday afternoon, so timing played its part as well.
Same race, different weather
Asked about the race, the skipper of the winning Class 40 said they had only an hour without wind. A class winner further back called it a race where boats 20 or 30 miles apart can have completely different weather. The tracker agrees.
Hours spent under 3 knots, by class (median boat)
From the race tracker, 5-minute positions, start to finish. The fastest quarter of the fleet spent a median 2.9 hours under 3 knots, and the slowest quarter 15.8 hours.
The fastest boats were through before the holes off Catania and north-west Sicily set in, while the smallest spent most of a day going nowhere. That’s timing, and it’s why we judge routes by sailing them through the same weather rather than comparing finishing times.
The races within the race
Matched boats show what the forecast can’t, the choices crews made on the water. We describe boats by model, not name.
Two identical Pronavia 38s, one sailing school
Same boat, same school, two skippers. Level on Saturday afternoon but 6 h 21 apart at the finish on corrected time. The gap grew whenever one boat found the breeze first, above all getting out of the hole off Trapani was the big difference. The boat behind took back 14 NM on Tuesday night on our conservative line, then lost it again on the beat to Lampedusa.
Lead in miles to the finish: Sat 15:00Z: 0 NM · Sun 09:00Z: 12 NM · Mon 09:00Z: 17 NM · Tue 15:00Z: 36 NM · Wed 05:45Z: 22 NM · Wed 12:00Z: 30 NM
Headed south of Pantelleria: hold on, or tack early?
A shift on Tuesday afternoon headed the boats sailing south from Pantelleria. The IMX 40 held on and tacked twice whereas the Oceanis 46.1 tacked early and four times and the First 36 tacked twice but was furthest back. By the next morning the IMX 40 had gained about 9 NM on the Oceanis and 23 NM on the First 36. Crew experience or luck?
Two Sun Fast 3600s, in different classes
The same design, 16 h 50 apart on elapsed time. One stayed near its class median all race, while the other was near the back of its class from the first night, and lost most in the hole off Trapani. Same hull shape, the difference is crew, sails and where they sailed.
Two First 40.7s: different crews, different challenges
One boat stopped twice to fix problems, for about 12 hours in all, and still finished. 7 h 57 apart on corrected time but less than its time lost to repairs.
What we’re taking into the next race
- Coastal detail matters, and a sharper model can see it. On leg 1, ICON-EU’s weather put the best route where the fleet went, but our coarser forecast didn’t. That makes the case for using the sharper model near land in our routing, where it reaches, and it’s next on our list.
- Score each decision on its own. The winners mixed our options, going inside on leg 1, north at the Aeolians and east at Lampedusa. Our next briefings will show the count on each side of every decision, leg by leg.
- Time the conditions to when you get there. The back of the fleet met different weather on every leg from the boats a day ahead. Conditions should follow when the boat actually gets there.
- Plan for a wider range than the forecast alone. Boats rated like a First 40 finished across 96 to 129 hours. The weather explains part of that, and crew, sails and luck the rest.
- The biggest gains came from decisions on the water. When to tack after a shift, when to leave the coast. That’s the moment our on-board app is built for.
How we worked it out
- Our briefing: ECMWF IFS ensemble, 50 members, 00:00 UTC run of 16 October 2026, re-run for the 11:10 CEST start. First 40 (ORC certificate polar) at 92%. Copernicus Marine currents. The published file and the engine that made it are archived, so it re-runs exactly.
- The weather that actually happened comes from ECMWF HRES analyses (the first hours of each new run, stitched every 3 hours, used at 0.25°, about 15 NM), and DWD ICON-EU (the first hours of each run, hourly, 0.0625°, about 4 NM). ECMWF open data CC BY 4.0, and DWD open data.
- The fleet: YB Tracking positions every 5 minutes, joined to the official IRC results. We publish only counts, medians and anonymised pairs, never individual tracks.
- Class percentile: a boat’s place in its IRC class on corrected time, scaled from 0 (winner) to 100 (last finisher). Light air: time between 5-minute positions with speed over the ground under 3 knots.
- Not modelled: tacking and gybing losses, sail choice, and wind shadows finer than each model’s grid.
ProSail compares route options across forecasts to support the skipper’s judgement. It doesn’t promise a safe or a winning route.