Every spring, the same conversation happens in every boathouse in the country. Someone says their crew would beat someone else's crew. Someone disagrees. There is no obvious way to settle it, because the two crews live a thousand miles apart and have spent the season racing entirely different sets of opponents.
The Rowing Power Index is one attempt to settle a few of those conversations honestly. It is a model. It reads every result we have, on both sides of the Atlantic, and produces an RPI for every crew that has raced enough to be rated.
One number is a polite fiction
Most ranking systems give you a single number per crew and call it a day. Single numbers are easy to read. They are also slightly dishonest in a sport where the same crew can win one weekend and finish fourth the next, and where two crews you'd love to compare may not have been on the same body of water all season.
The RPI gives you a four-digit rating, anchored around 1500, plus an interval around it that says how confident the model is in the number. The men's V8+ leader on the morning of April 28, 2026 is Greenwich, RPI 2002, with a 90% credible interval running from 1903 to 2099. Sitting at #2 is St. Joseph's Prep, RPI 1946, interval 1782 to 2112. The two intervals overlap almost entirely. The model is telling you that Greenwich is its best guess for the top crew and that the gap to St. Joe's is, statistically, not yet a gap.
The interval is the part most rankings won't show you. It does most of the work on this site.
Why some crews aren't shown at all
The public rankings only include crews with at least three rated races this season. A model trained on one data point is just gossip with a number on it.
This matters most for the spring scholastic landscape. The strongest UK schoolboy crews, including Shiplake College and St. Paul's, raced once at the Schools' Head of the River in March and have not raced enough this season to qualify for the public table. Their numbers exist on their team pages, with the appropriate caveats. They will lift into the rankings when the racing earns it.
When a crew has an RPI but no rank — visible on a team page or a matchup screen — that is the model telling you it has a guess, but not yet a defensible answer.
How the rating moves
Every race the model scores adjusts ratings two ways at once.
The first is local: a win or a loss against a particular opponent moves the two crews relative to each other, weighted by margin, conditions, and the importance of the event. A canvas-margin win against a strong field at Stotesbury or the SRAA carries more than open-water at a local invitational.
The second is structural: the model also estimates each league's mean strength alongside each crew's rating, jointly, from the same data. When Atlanta Junior beats St. Andrews RC at Dogwood, the result moves both crews and very slightly nudges the model's belief about Southeast strength. When Greenwich wins the King's Crown A Final, the model adjusts its read on the Northeast at the top.
That second move is what makes a Greenwich win at a Northeast invitational meaningful even when the Southwest favourites aren't on the start line. The model is filling in the cross-region gaps the calendar leaves behind.
Head-to-Head Record
Most-raced opponents — win/loss based on finishing position in shared races
| Opponent | W | L | Record |
|---|---|---|---|
| Row America Rye | 6 | 1 | 86% |
| Community Rowing | 7 | 0 | 100% |
| St. Joseph's Prep | 4 | 3 | 57% |
| Saugatuck | 6 | 0 | 100% |
| PNRA/Mercer | 6 | 0 | 100% |
| Central Catholic | 6 | 0 | 100% |
| D.C. National | 4 | 0 | 100% |
| Pelham Community Rowing Association (PCRA) | 4 | 0 | 100% |
| Duxbury | 4 | 0 | 100% |
| Gonzaga | 4 | 0 | 100% |
Greenwich vs most-faced opponents across all recorded races
Calibration, not accuracy
Across the last several months of completed races we have checked our work on, the model called the winner roughly nine times in ten when both crews were in the database. The headline is fine. The texture matters more.
The number that actually counts is calibration: when we say a crew has a 70% chance of winning, do they win 70% of the time? Mostly, yes — and the methodology page publishes the calibration plot so you can see where it sags. The model is, if anything, slightly overconfident in the 55–70% band. We list that on the page as a known weakness.
How to read the page
A crew at the top of a ranking with a tight interval is the model's strongest statement: it has seen a lot of racing and is confident in roughly where the crew sits. Winter Park, currently #1 in the women's V8+ at RPI 2191 with an interval just over 200 points wide, is the cleanest example.
A crew with a wide interval is the model reserving judgment. The width of the bar matters as much as where the bar sits. St. Joseph's Prep's interval runs more than 300 points top-to-bottom, which is why the model has them second on the rating and not stamping it with confidence.
Two crews with overlapping intervals are, statistically, candidates for the same position. The rankings tell you who's in the conversation; the head-to-head tool, the championship simulators, and your own eye decide who wins on a given Saturday.
Welcome to the Power 10. Let's row.