A world best time takes a fast crew and a fast day. For each elite record I estimated how much came from each. I compared the record with a typical Worlds gold medallist from the same year, and with how fast every other event went on the same day. That gives the record's margin over today's typical gold medallist, and how often a crew like that would beat it. Further down are the unlucky and the great crews, next season's chances, how records are paced, and the countries, lakes, winds and times of day behind them.
The data covers every race at the Worlds, Olympics, World Cups, Europeans and final Olympic qualifiers since 1985: . All times, records and rankings are correct as of . New records set after that date are not included, so parts of this page will go out of date.
A typical Worlds gold medallist in 2026 would beat the W4-, M1x or W1x record in only 1 A-final in 500 to 750. At about five A-finals a year, that works out at more than a century. Treat that as a way of saying "very unlikely to fall soon" rather than a forecast, because boats, training and the depth of each event will change long before then. The M2- is next, at about 1 in 200.
Murray and Bond set 6:08.50 in a heat at London 2012, on a day that was faster than average but not unusually so (75th percentile). Allowing for the conditions and for racing in a heat, they were about 4.0% faster than a typical 2012 Worlds gold medallist. That is the largest crew margin behind any current record.
Both were set on days among the fastest 4%, by crews that were no faster than a typical Worlds gold medallist once the conditions are allowed for. A typical gold crew would beat them in about 1 A-final in 14 (M4x) and 1 in 17 (M8+), so both are likely to fall within about four years.
A New Zealand pair finished 0.54 s outside the M2- record at World Cup II in 2026. Crews have also come within 0.2% of the M2x and W2x records. Even a hard record can be nearly matched when a strong crew gets a fast day.
Typical 2026 gold is the time the model expects to win a Worlds A-final in the class, in average conditions, in 2026. Gap to beat is how much faster than that the record is. The chance per A-final combines that gap with conditions sampled from 505 real race days and the normal race-to-race variation of winning crews. It is calibrated so that it predicts the number of records actually broken in A-finals since 2005. Hard means less than 1 in 100 finals, and soft means more than 1 in 40.
| Class | World best | Age (yrs) | Typical 2026 gold | Gap to beat | Chance per A-final | Expected wait | Closest since |
|---|
Gap to beat is shown as % of speed. Expected wait = 1 ÷ chance per A-final ÷ A-finals per year in the class (2023–25 average). Retired classes have been raced too rarely since 2000 to model.
* Waits of more than 100 years are an extrapolation of today's odds, not a prediction. Over that long, equipment, training and the depth of each event will change, and so will the odds, so these records are very unlikely to last that long.
Each bar shows how much faster the record was than a typical Worlds gold medallist of the same year, split into three parts:
Whichever part you sort by is drawn from zero, so those lengths can be compared directly, and the other parts stack after it. When a part is negative, such as a crew slower than a typical gold medallist, it steps back to the left as a hatched block over the rest of the bar. So the far end of the coloured bar is not always the total. The black tick always marks the total margin.
The right-hand column shows the value being sorted on. For crew and conditions it also gives how fast the day was as a percentile, where 100th is the fastest of the 505 race days. On speed, the middle 90% of race days run from to against average, and the fastest 1% are at least . Most current records were set on days in the top few percent: Amsterdam 2014 (seven records) and 2026, Tokyo 2021 and Seville 2002. Too few events raced on the LM1x record day to estimate its conditions, so it is not shown. Classes in grey were not raced at the 2025 or 2026 Worlds.
Each row is one boat class, and each bar is a period during which one time stood as the fastest, rebuilt from the race data. Hover a bar for the crew and how long it stood. The rebuilt histories finish on the official world best time in all 26 classes. The pale first bar in each row is the first time in the data, not a record set then.
Each dot is an A-final winning time, as % of the current record's speed (100% equals the record). Orange-ringed dots mark every race that set a new best time, one at each corner of the line, coloured by regatta like the other dots. The dashed first ring is just the first time in the data. Some of those were heats or semi-finals rather than A-finals (the men’s pair record was set in a heat). Every result comes from a Worlds, Olympics, World Cup, Europeans or final Olympic qualifier, so the four colours cover them all. Winners slower than the bottom of the axis are pinned to it. Hover any dot for the crew, that day's conditions and the crew's margin over a typical Worlds gold medallist.
Every A-final winner at the Worlds, Olympics and World Cups from 2022 to 2026, as % of world-best speed. The open ring is the class median and the orange dot is the fastest winner. A dot at 100% is a record set in this period. Winners below 92% are pinned to the left edge. Hover any dot for the crew and regatta.
Some crews never set a world best but were quick enough to on the right day. A crew's best-three rating is its average margin over a typical Worlds gold medallist in its three best A-final or semi-final results, with conditions and round taken out (the same rating as section 7). Only days no slower than −4% count, because the model fits the very slowest days worst. The day needed is how fast a race day the crew would have needed, rowing at that level, to break the record standing at the time. It is ranked among all 505 race days, so a 65th-percentile day means about one race day in three would have been fast enough.
Kenia Lechuga Alanis tops the list. Rowing at her best-three level, she would have broken Imogen Grant's LW1x record of 7:23.36 on about one race day in three. Lightweight crews take most of the top places, because their events are raced at fewer regattas and good crews get fewer goes at a fast day. Among open classes, Haigh and Scown (New Zealand W2-, 2011) needed only an 84th-percentile day to beat Romania's 2002 record. Two M2- crews were each one good tailwind from the pair record: Reed and Hodge against Pinsent and Cracknell's 6:14.27, and Chardin and Mortelette against Murray and Bond's 6:08.50. Of crews, would have broken the record standing in their time on a top-5% day.
| Class | Crew | Best time | Record at the time | Day needed | On a top-5% day |
|---|
Best time is the crew's fastest actual result. "On a top-5% day" is the time its best-three level would give on a day as fast as the fastest 5% of race days (+), with the margin it would have had over the record standing at the time. Crews that set any record are left out. Results from 2000 on, since the conditions model starts there.
Each crew's results since 2000 are measured two ways:
Pairs, doubles and singles are tracked as the same people across seasons. Fours and eights change seats, so each is one nation's crew for one season.
Murray and Bond lead on both measures. They won all 24 of their A-finals in the data, by an average of 1.3%, and their best-three rating of +4.9% is the highest of any crew in any class. Chardin and Mortelette, the French pair racing in the same years, are second overall (+4.0%).
A record needs a strong crew and a fast day at the same time, so record holders are not always the best crews. Robbie Manson's M1x record shows how that works. His 6:30.74 at the 2017 World Cup in Poznań was 6.9% faster than a typical 2017 Worlds gold medallist. About 4.8 points of that came from the day, which was among the fastest 2% of race days, with a tailwind of about 3.7 m/s. The other 2.1 points were Manson himself, and that race is his best result in the data. His best-three rating, the average of that race and his next two best, is +1.4%. That ranks him 20th of the 34 single scullers with at least five A-finals since 2000. The list is nearly every leading sculler of the era, so 20th is still very good, but it means the record owes more to Poznań's tailwind than to Manson being the fastest sculler of his time. Mahé Drysdale leads the list at +3.5%. The eights tell a similar story: the M8+ record crew (Netherlands, 2026) ranks 39th of 100 M8+ crews, while the W8+ record crew (Romania, 2026) ranks 5th of 95.
| Class | Best crew (best-three rating) | Rating | A-finals won | Record crew | Record crew's rank |
|---|
Rating is shown as % faster than a typical Worlds gold medallist. "Record crew's rank" is where the crew holding the current world best sits on the same rating, among all crews in that class since 2000. Crews need at least five A-finals (three for a four or eight season). Some crews race far more often than others, which gives them more chances at three great days, so compare crews within a class rather than across classes.
This takes the three best crews of 2026 in each class and their 2026 form, meaning their average margin over a typical Worlds gold in A-finals and semi-finals, conditions removed. Form is shrunk about back toward typical, because that's how much a crew's form carries over from one season to the next on average. The model then plays out about five A-finals with conditions drawn from real race days. The chance shown is that at least one of those crews breaks the record at least once.
Each line shows a race's speed over each 500 m as % of its average speed for the whole race. The band is the middle half of A-final winners since 2001 in the classes raced today, and the dark line is their median. A typical winner goes out about above race pace, is slowest in the third 500, and sprints back to about average. Records follow nearly the same shape. Their median first 500 is above race pace and their last 500 is . Only of winners, and of record setters, row the second 1000 faster than the first.
Great Britain and Italy have set the most elite records since 1985, with 17 each, and New Zealand holds the most current elite world best times, with four. Per head of population, New Zealand is far ahead: its four elite world bests work out at 7.5 per 10 million people, almost three times Croatia, the next best. Switch the measure below.
All levels adds the U23 and U19 world best times. Records set counts every new elite best time in the rebuilt histories (section 3), skipping each class's first time in the data. Years holding adds up how long each of those stood. East Germany (RDA), West Germany (RFA) and the Soviet Union (URS) appear only in the historical measures. Populations are approximate 2025 figures. Mixed events are left out.
Each dot is one race day since 2000 at the venue, placed by its conditions effect, which is how much faster or slower every event ran that day than usual. The open ring is the venue's average. Varese and Plovdiv have the fastest average days, and Lucerne is reliably quick across 92 race days. The Bosbaan in Amsterdam is the lottery. Its average day is slightly slow, yet more than a quarter of its days are among the fastest 10%, and it holds 10 of the 26 current elite world best times.
Venues with at least 8 race days since 2000 are shown, and days below −12% are pinned to the left edge. A fast day is one in the fastest 10% of all 505 race days (at least on speed). The effect mixes the course itself with the weather it tends to get during the racing season, so a tailwind lake and a fast course look the same here. Right-hand column: share of fast days · current elite world bests set there.
I matched every race since 2000 to hourly ERA5 weather at the course, from the Open-Meteo archive. The data shows which wind direction makes each venue fast, which gives the course direction without needing a map. To keep that honest, each day's tailwind is measured along a direction learnt from the other half of the years. Each point below is one race day.
Across race days, wind along the course is most of what separates a fast day from a slow one, but the relationship is not a straight line. With separate slopes either side of still air, each 1 m/s of headwind cost about , while each 1 m/s of tailwind gave only about . A headwind hurts about three times as much as the same tailwind helps, which fits air drag rising with the square of the boat's speed through the air. A smooth curve (quadratic) shows the same flattening, and both fit clearly better than a straight line (). Crosswind made no clear difference. Each 1 °C of warmer air added about . Within a single day, races rowed in a stronger tailwind than earlier or later races were faster by per m/s. The 2014 Worlds in Amsterdam, with seven current records, had tailwinds of about 7 m/s during the finals. Murray and Bond's M2- heat in London had 2.0 m/s.
| Venue | Days | Fast wind from | Per m/s | Typical wind |
|---|
Tailwind is the wind component along each venue's fast direction, averaged over the hours that day's races ran. Only venues with at least 12 race days are included. "Per m/s" is that venue's own fit, which is noisy for venues with few days or little wind (Varese, Bled). Weather is from a 0.25° reanalysis grid, not the course itself, so it is smoother than what the crews felt.
This compares winners' speed with other races on the same day at the same regatta, after allowing for class and round. Winners are fastest around 11:00 to 12:00. By 15:00 to 16:00 they are about 0.5% slower, as the average wind at these regattas rises from 2.2 m/s in the morning to 2.9 m/s. A stronger wind usually means rougher water, whatever its direction. Races at 9:00 are also slightly slow, in cooler air (about 18 °C against 23 °C mid-afternoon).
Local start times. Before 2005 the API has no start times, and a few regattas whose times don't look local are left out. Based on winning times.
Without race data, the only evidence is the records themselves. Boat speed should rise with crew size, so I fitted log speed ~ log rowers + women + lightweight + sculling + cox to the 26 records. Crew size gets an exponent of (theory gives 1/9 ≈ 0.111), with R² . Each bar shows how far a record sits above or below that line, with the class left out of its own fit.
This view cannot separate a strong crew from a fast day. The cox term is also fitted mostly on the old coxed fours and pairs, which flatters both eights. With race data, the M8+ comes out as one of the softest records (section 1).