Every year, a few weeks before the Boat Race, Oxford and Cambridge used to line up their crews in front of photographers and step onto the scales. The weigh-in was a tradition stretching back to at least 1874, when the average rower tipped the scales at just under twelve stone. For well over a century it was part of the ritual: the ceremonial squaring-off, the comparing of physical statistics, the first competitive intelligence of the season made public.
That tradition has quietly been unwound. In 2019 the women’s crews were weighed for the last time. By 2022 the women had formally opted out, and even for the men the published data had been stripped back. By 2025, the official Boat Race article on the crew announcement contained exactly one sentence on the subject: “Of the two men’s Blue Boats, Cambridge was the heavier. The women opted not to be weighed in.” No averages. No individual figures. Just a comparative direction of travel.
For 2026, today’s race, there is nothing at all in the public record.
The official explanation, offered by rowing historian Tim Koch at Hear The Boat Sing, is that the weigh-in had become “mildly undignified”, a cattle-market atmosphere that athletes, particularly women athletes, were understandably reluctant to submit to. That is entirely credible. But there is a sharper possible explanation, and it has a date attached to it: 2020.
The formula
In 2020, academics Rutger Lit and Siem Jan Koopman at Vrije Universiteit Amsterdam published a paper demonstrating their time series modelling software. Their test case was the Oxford-Cambridge Boat Race. They fed 191 years of race results, every contest from 1829 to 2019, into a statistical model and asked a simple question: what actually predicts who wins?
They tested three variables available just before each race: the outcome of the coin toss, the margin of the previous year’s victory, and the log difference in average crew weight between the two boats.
The result was unambiguous. The coin toss? Statistically insignificant, effectively a coin flip, which of course it is. Previous winning margin? Also insignificant. Average crew weight? Statistically significant at p < 0.01, with a coefficient of 18.214.
In plain English: across 191 years of races, the heavier crew has won often enough, and consistently enough, that there is less than a 1-in-100 chance this is down to luck. The coefficient of 18.214 tells you how much weight moves the needle: dial in a 5kg-per-man Cambridge advantage and their probability of winning shifts from roughly 67% to around 83%. Dial it the other way, with Oxford 5kg heavier, and Cambridge go from favourites to slight underdogs.
No other variable in the model (not the coin toss, not last year’s winning margin, not which station a crew takes) comes close to that kind of predictive power. Weight, in a race between athletes at the absolute peak of their conditioning, turns out to be the thing that matters most.
“The heavier boat is faster. This can be explained by the fact that heavier rowers have on average more muscle mass and can have a stronger pull.” — Lit & Koopman (2020), Vrije Universiteit Amsterdam
The model also discovered something else: the Boat Race runs on a cycle. Oxford and Cambridge don’t simply trade wins at random. The statistical model finds that dominance tends to shift roughly every 12 years — one university builds momentum, holds it for a decade or so, then gradually cedes ground to the other. Cambridge’s current run of six wins in seven years is entirely consistent with where the cycle says we should be.
What the model contains:
| Variable | Significance | Effect |
|---|---|---|
| Crew weight difference | p < 0.01 — significant | ±16 percentage points per 5kg |
| Coin toss winner | p = 0.76 — not significant | ±2.8 percentage points |
| Previous winning margin | p = 0.73 — not significant | Negligible |
Important caveat: The model covers men’s races only. It was built on data beginning in 1829; no equivalent academic model exists for the Women’s Boat Race.
How accurate has it been since?
The paper was published just as the 2020 race was cancelled due to Covid. Lit and Koopman noted, wryly, that without weight data for the projected 2020 crews, they could only estimate an 80% probability of a Cambridge win. We will never know if they were right.
What we can test is the model’s performance since 2021:
| Year | Model prediction | Actual winner | Correct? | Notes |
|---|---|---|---|---|
| 2021 | Cambridge | Cambridge | ✅ | — |
| 2022 | Cambridge | Oxford | ❌ | Oxford were heavier; no weight data available in forecast mode |
| 2023 | Cambridge | Cambridge | ✅ | Cambridge 89.2kg avg vs Oxford 92.0kg — weight favoured Oxford, cycle won |
| 2024 | Cambridge | Cambridge | ✅ | Near-identical weights (91.65 vs 92.23kg); cycle decisive |
| 2025 | Cambridge | Cambridge | ✅ | Cambridge heavier per official statement; no figures published |
| 2026 | Cambridge | ? | — | No weight data published |
Four correct from five: 80% accuracy across a genuinely post-sample period. The one miss, 2022, is the most instructive: Oxford were the heavier crew that year, and Oxford won. Had we known the weights, the model would very likely have flipped its prediction. The miss happened precisely in the data vacuum that has since grown larger.
The coin toss problem
This is worth dwelling on, because the coin toss is one of the great theatrical moments of Boat Race Day. The two presidents gather on the bank. The toss determines which station, Surrey or Middlesex, each crew will take. The station choice matters enormously for race tactics and for where a crew can make a move.
And yet: according to 191 years of data, it predicts the winner of the race with essentially zero reliability. The coin toss winner’s contribution to the model is −0.11, against the weight variable’s 18.21.
Knowing who won the coin toss shifts the probability of winning the race by roughly ±2.8 percentage points. A 5kg weight advantage shifts it by ±16 percentage points.
The coin toss is theatre. Weight is substance. And weight is the figure we are no longer given.
To put the numbers in context: elite Boat Race rowers typically weigh between 90 and 95 kilograms. A three-kilogram-per-rower advantage – the difference between Oxford averaging 93.5kg and Cambridge averaging 90.5kg – sounds small. It is roughly one decent meal per man.
But the model says that three kilograms per rower is enough to cancel Cambridge’s entire 67-33 cycle advantage and reduce the race to a coin flip. That is what the weight variable is actually measuring: not bulk, but the muscle mass that translates directly into pulling power over 6.8 kilometres of tidal Thames.
The interactive prediction engine
The widget above uses the Lit & Koopman (2020) model parameters (φ₁=1.6412, φ₂=−0.8976, β=18.214) to show Cambridge’s probability of winning the men’s race. Because no weight data has been published for 2026, the weight slider opens at zero. Use it to explore the full range of possible weight scenarios.
Why does the withdrawal matter?
Consider the timing. The Lit and Koopman paper appeared in 2020. It was immediately and freely available online. It identified, precisely and publicly, that crew weight was the single variable that could shift a Boat Race prediction by up to 16 percentage points.
The bookmakers take the Boat Race seriously. Today’s men’s race has Cambridge priced between 1/6 and 1/7 across major bookmakers, implying a probability of around 85–87%. Here is how the odds compare:
| Bookmaker | Cambridge | Oxford | Cambridge implied probability |
|---|---|---|---|
| William Hill | 1/6 | 7/2 | 85.7% |
| Paddy Power | 1/7 | 7/2 | 87.5% |
| Sky Bet (early market) | 3/10 | — | 76.9% |
| Freetips.com | 1/8 | — | 88.9% |
Women’s race: Oxford 2/5 (71% implied) · Cambridge 7/4 (36%)
Note that the men’s market opened weeks ago at 3/10 and has tightened to 1/6–1/7 as race day approached, driven by fixture results and squad intelligence rather than any published weight data.
We are not suggesting that the Boat Race Company withdrew weight data because of a Dutch academic paper. The dignity argument is real and legitimate, and the women’s opt-out predates the paper. But it is a striking coincidence that the only publicly documented formula for exploiting weight data in Boat Race prediction emerged just as weight data began its disappearing act.
So who will win today?
The model (running on cycle data alone, with no weight information) gives Cambridge approximately a 67% probability of winning the men’s race. The bookmakers, drawing on fixture results, squad composition, and whatever intelligence they have gathered across the season, are pricing Cambridge at 1/6 to 1/7, implying 85–87%.
That 20-percentage-point gap is precisely where the weight variable lives. Whether it reflects a genuine Cambridge weight advantage, or simply their visible dominance in pre-race fixtures, or their coaching continuity under Rob Baker, we cannot know, because the weight figures have not been published.
What we can say is that if you want to apply the Lit and Koopman model in full (the way it was designed), you need data that is no longer public. The model’s best estimate without that data is 67% for Cambridge. If you’re a Cambridge supporter, the market’s 86% is probably reassuring. If you’re an Oxford supporter, that gap is the space in which hope still lives.
The women’s race: a different story entirely
The Lit and Koopman model covers only men’s races. There is no equivalent statistical model for the Women’s Boat Race, and given that women’s weights have not been published since 2019, there is not enough public data to build one.
What the bookmakers tell us is this: Oxford Women are favourites at 2/5 (implying 71% probability), despite not having won the Women’s Boat Race since 2016 and despite Cambridge winning nine in a row. That is a striking assessment. It reflects the quality of Oxford’s crew this year, anchored by Olympic bronze medallist Heidi Long as stroke, and the view that the gap between the two squads has finally, genuinely closed.
Cambridge women have won nine consecutive races. And yet this year, on the basis of the squad assembled and the season’s evidence, the market makes Oxford slight favourites. If they win today, it will be the kind of result that no model (because no model exists) could have predicted with confidence. Which is perhaps the most honest thing that can be said about prediction: sometimes the data runs out, and the race just happens.
What today would tell us
If Cambridge win today — as the model and the bookmakers both expect — that proves nothing in isolation. The real test comes when the cycle turns, probably within the next two to four years. The question then will be whether weight data, had it been available, would have given advance notice.
The race starts at 3:21pm. The crews know how heavy they are. We don’t.
Sources & methodology: Prediction model: Lit, R. and Koopman, S.J. (2020), “Forecasting the 2020 edition of the Boat Race“, Vrije Universiteit Amsterdam / Time Series Lab. Model parameters: φ₁=1.6412, φ₂=−0.8976, β(weight)=18.214. Backtesting conducted by Putney.news across 2021–2025 race results. Odds sourced from Paddy Power, William Hill and Sky Bet, 4 April 2026. Historical weigh-in data from heartheboatsing.com (Tim Koch), University of Cambridge press releases 2012–2024, and the official Boat Race Company. The prediction widget is for editorial illustration only. Crew weight data for 2026 is not publicly available; the widget’s zero baseline reflects this absence.
A fascinating story showing how data-driven predictions can bring new understanding to sporting events. Curious that the weights of rowers have been removed, given how this information could affect betting results. It suggests a strong commercial incentive to employ subterfuge to get hold of rowers’ weights.