How ThePower Score Works

Last updated: Oct 20, 2026 (list 1) · Next update: Oct 21, 2026, results check for SUP World Cup Suzhou, ICF SUP World Championships 2026 (long distance results); a new list publishes when results arrive. Method v1.3.

The Short Version

Every race is a set of head-to-heads. You get credit for everyone you beat, more for beating stronger paddlers and more for beating them by more. Your rating is the number that best explains all your head-to-heads in the last two years, on a scale where a 100-point gap means the higher paddler finishes ahead about two times in three, and a 200-point gap about three times in four.

We never compare a time from one race with a time from another. Courses, distances, wind and current change from race to race, so a 1:02 in one place says nothing about a 1:02 somewhere else. Inside one race, everyone shared the same water on the same day, so that is the only place we compare.

What The Number Means

Every paddler starts at 1000. It is the anchor of the scale, not a grade: each paddler carries 1 virtual win and 1 virtual loss against a 1000 paddler, so someone with one race sits close to 1000 until more races say otherwise. Real results push the number up or down from there, and it can fall below 1000.

A rating only means something next to another rating. Here is what a gap between two paddlers in the same table says:

So a paddler at 2000 against one at 1500 is a 500-point gap: the 2000 paddler finishes ahead 95 % of the time, about 19 races in 20. Two paddlers at 1500 and 1550 are close to a coin flip, 57 % for the higher one. A paddler at 1500 is not "half as good" as one at 2000; the gap is what carries the meaning, the same gap anywhere on the scale.

A rating belongs to one table: one discipline, one gender, one board class and one age group (Open or a Masters group). A Distance 1700 and a Sprint 1700 are different things, and so are an Open 1700 and a Masters 50+ 1700: those paddlers never met in the same fit. Ratings are shown as whole numbers.

How One Race Moves It

Each pair of finishers in a race counts as one head-to-head, so a race of 27 finishers holds 351 of them, 26 for each paddler. A head-to-head is not just a win or a loss. The gap between the two is measured as a percent of the reference time, the mean of the 3 fastest finishers, so a short race and a long one speak the same unit. The gap then becomes a share of a win:

• A 1 % gap counts as a 73 % win, a 3 % gap as 95 %, and a dead heat as 50 %.
• The curve flattens out: a 10 % gap and a 20 % gap both count as a near-certain loss, so a back half that stretches out when the wind dies changes little.
• Paddlers who finish together in a drafting pack, with the same time, split the head-to-head 50/50. So does a place the results print as shared.
• Where a race has no times, only places, a better place counts as a full win.

Who you beat matters as much as by how much. Beating a 1400 paddler by 1 % is worth more than beating a 1000 paddler by 5 %, because the fit already knows who they are. Losing to stronger paddlers costs little; winning a weak field gives little.

A Real Race, Worked Through

The Open women's long distance final at the ICF SUP World Championships 2025 in Abu Dhabi, about 6.2 mi (10 km): 27 finishers, every one of them timed. The reference time is the mean of the top three, 1:02:07.6. Here is how the winner's head-to-heads read against a few of the field:

Second place finished 0:03.3 back, 0.09 % of the reference time, so that head-to-head counts as a 52 % win for the winner: close to a dead heat, which is what it was. 0:21.5 to fifth is already a 64 % win, 2:07.6 to seventh 97 %, and the last finisher, 14:40.2 back, counts as a full loss. A negative figure behind the top three means faster than their mean; that column is the per-race score each athlete page prints beside a result.

What release 1 made of it. These numbers are from release 1, the first list (Oct 20, 2026); later lists move them as new races come in.

• The winner's Distance rating is 1731 ± 94, provisional after 2 races. It reads 1000 + 706 from the 2024 World Championships long distance race, 7.1 mi (11.5 km), where she finished second of 54, + 25 from this win.
• Why only 25 for winning a world championship final? Races are added in date order, each time finding the rating that best fits the races so far. By the time this race is added, the 2024 race has already moved her from 1000 to near the top of the table. Against paddlers rated close to her, a win by 3.3 seconds is only a little more than that rating already predicted, so it added 25.
• The runner-up, a few seconds behind, gained 50 from the same race. What a race is worth depends on what your other races already say.
• Her 2024 race still counts, at 55 % weight: it is more than a year old, so it counts for less (see Form).
• Rated 1731 and 1681 in Distance, the two would line up again with the winner favoured at 57 %: a 50-point gap, close to a coin flip.
• Her headline Power Score is 2148: Sprint 1883 is her best established discipline, and Technical 1882 adds 0.3 × its 882 points above 1000. Distance, still provisional, does not count toward it yet.

Bands And Provisional Marks

• The ± band says how sure the number is. A paddler's true level sits within the band about two times in three. The more head-to-heads a paddler has, the narrower it gets: the winner above carries ± 94 after 2 races.
• Provisional means fewer than 3 counted races in the last two years. A race run in heats and finals, or in stages, counts once. A provisional discipline never leads the headline Power Score while the paddler has an established one.
• Regional marks a rating from a part of the world whose paddlers rarely race anywhere else, so it compares within that region only (Pools and bridges).

Predictions And Head-To-Head Odds

Two ratings in the same table give odds: the chance that one paddler finishes ahead of the other the next time they race. That is the whole prediction. We do not forecast places or times.

On every athlete page, Compare lets you pick another paddler and read three things:

• The odds, from the two ratings in the table you both share; where you share more than one, the one with the most races between you. Never from the headline Power Score, which carries a bonus for a second discipline.
• The record: who finished ahead in every race you both finished in the last two years, by time where both have one, else by place, and when you last met. Under 3 shared races the record gives no verdict.
• When they disagree, the page says so: "The ratings favour one paddler; the record favours the other." A record counts wins and losses; the ratings also weigh margins and everyone else each of you raced.

Do the odds hold? Before launch we tested the method on public results: trained on the 2023 and 2024 World Championships long distance races, it was asked about 2025. When it gave the favourite 90 % or more, the favourite finished ahead 174 times in 182 (96 %); at about 80 %, 28 times in 32 (88 %); at about 70 %, 15 times in 19 (79 %). Finals among near-equals stay close to coin flips in any system, which is why we show odds, not certainty.

Form

Recent races count more. A race counts in full for 365 days (12 months). After that its weight falls in a straight line to 0.5 at 730 days (24 months), and then it drops out, so a race from two seasons ago cannot decide this year's list.

• An athlete page shows each rating's form over the last 12 months: the running rating after each counted race as a small line, and one sentence in the form "Up N points over 12 months; best race …; 5 of 6 results count".
• The tables show each paddler's move against the list before: up or down with the points, level, or New.
• No streaks, badges or arrows that imply more than the numbers.

Five Disciplines And The Overall

A great sprinter is not automatically a great distance paddler, so each discipline is rated on its own. Every race is tagged to one discipline when its results come in; an editor confirms the tag, and it stays on the result.

• Sprint: up to about 270 yd (250 m), usually about 220 yd (200 m).
• Technical: about half a mile to 3.1 mi (800 m to 5 km) with buoy turns, beach starts and surf.
• Distance: 3.1 to 21.7 mi (5 to 35 km).
• Downwind and crossings: by format, whatever the distance: downwind runs and open-water crossings such as the Gorge, Molokai 2 Oahu and Koa Kai.
• Ultra: over 21.7 mi (35 km) in a day, or a stage race whose stages add up to over 21.7 mi. A race between 18.6 and 21.7 mi (30 and 35 km) goes to the format the organizer names.

Heats and finals. Sprint and Technical races often run in rounds: heats, quarter-finals, semi-finals, finals. Each round is its own start, so you are compared only with the paddlers in your heat or your final, who raced the same water at the same time. The race as a whole needs 10 paddlers over all its rounds, and a round needs 2.

Board classes. 14', 12'6" and Unlimited are each rated on their own. Results that print no class count as 14'. 14' is always a table; 12'6" and Unlimited become a table of their own once a discipline has 30 rated paddlers in that class, so the order means something. Until then the board filter says how many there are so far. Inflatables are recorded, never rated.

The headline Power Score, the Overall, starts from your best discipline, and a strong second discipline adds to it: best discipline rating + 0.3 × the second best's points above 1000.

• A second discipline only adds, never subtracts. Sprint 1600 and Distance 1300 make a Power Score of 1690; a pure sprinter at 1600 stays at 1600.
• A second discipline at or below 1000 adds nothing, and a third never counts.
• Only established disciplines count. One still provisional (fewer than 3 counted races) is never the best while you have an established one, and never adds as the second, so one or two races never become a headline score.
• With no established discipline at all, the Power Score is your best provisional rating, marked provisional.
• With no established second discipline, the Power Score is your best rating alone, marked "one discipline".
• The headline Power Score and the nation table read the 14' Open ratings.

Open And Masters

The world tables, the Power Score, the nation table and a city's Top 10 are rated from Open races only. Each Masters group (40+, 50+ and up) is rated on its own, with its own table under the age filter, and a paddler who races only Masters is not in the Open table. A paddler who races Open and Masters has a rating in each.

The reason: at a championship almost nobody races two age groups, so rated together, a Masters 50+ win would read like an Open win. Junior races are rated on their own too, and never published as a table (Juniors).

What We Count, And What We Never Count

Coverage today. The first list, release 1 of Oct 20, 2026, read 7,520 results from the 2024, 2025 and 2026 ICF SUP World Championships and counted 7,035 of them; the rest did not finish or fell outside the rules below. That is a narrow window on the sport for now: a paddler who races mostly outside the World Championships has fewer races here than they raced. We add races as organizers, timing providers and federations send results, and every source we count is listed below.

• A start is one race for one gender, age group and board class. Paddlers in different categories or classes are never compared, even on the same course.
• Waves that started apart are separate starts.
• A start needs at least 10 finishers.
• A DNF, DNS or DSQ never enters a head-to-head and never costs rating. Relays and team boats do not count.

Where results come from. A file from the organizer or the timing provider, sent from the race's event page; an organizer's results page, with permission; a federation's published results; or a timing provider's API, by agreement. We never copy results from PaddleGuru, Webscorer, Athlinks, RunSignup or RaceSplitter: each forbids it in writing, and our system refuses them before it fetches anything.

What never counts. No editor-set importance, no prize money, no media factor, no bonus for events Supconnect covers or a Host runs, and no points from other rankings. Only finishes.

Results Sources

Organize or time a race? Send us your results from the event's page. A correction or a question about the method: info@supconnect.com.

Releases, Updates And Corrections

Ratings are published as numbered lists, and every list has a link that never changes. A list goes live as soon as it is computed, and checks run on every list to tell us when something needs a second look. A list found wrong is withdrawn and the one before it comes back. Every list stays on record: all releases.

When lists update: we check for new results every morning. A new list publishes only when the results it counts have changed, and on the 1st of every month, when the two-year weighting moves on. After a major race (a ICF, ISA, APP, ESL, EUROTOUR race, a featured race or a Pro race) we look for its results every day for 21 days.

A paddler who claims their athlete record sets their flag within one rule and may pick a home city from our city hubs; neither ever changes a rating. Spot a wrong result? Ask for a correction from the athlete page. Every flag change is on the flag-change list and every correction on the corrections log.

Juniors

A junior's races are real races, so they count in everyone's rating, theirs included. But no junior is named anywhere in the Power Score: not in a table, the nation table, search or an event's results, and a junior has no athlete page. Where a junior raced, their line reads "Junior", and ranks count adults only. A paddler is a junior when the results give a birth year that does not make them 18 for certain (born in or after the year that turns 18 this year); where records disagree, the youngest birth year counts. A birth year that makes them 18 or older on every date this year settles it: they are an adult, whatever age groups they raced in. Without a birth year, a paddler is a junior while they have raced in a junior age group (Junior, Youth, Cadet, U16 and the like) in the last 24 months, the same window the ratings use; older junior results no longer count. A paddler whose age the results do not give counts as an adult, so a former junior is named once they age out.

Questions Paddlers Ask

For The Curious

The technical detail, for anyone who wants to check our work. Every number below is read from the code that computes the lists.

The Head-To-Head Formula

For two finishers of the same start, the gap is the time between them as a percent of the reference time, the mean of the 3 fastest times. The faster paddler's share of the head-to-head is:

outcome = 1 / (1 + e^(−1 × gap %))

The slower paddler gets the rest. Equal times, or a place printed as shared, give 0.5. Where either has no time, places decide: 1, 0, or 0.5 for a shared place. A race counts with weight 1 for 365 days, then down in a straight line to 0.5 at 730 days, then 0.

The Fit And The Scale

All the weighted head-to-heads of one discipline, gender, board class and age group go into one Bradley-Terry fit: the set of strengths that best explains every result at once. We fit it with the MM algorithm, which treats every race at the same time, so the result never depends on the order races came in, and the same results always give exactly the same numbers.

The prior: every paddler also gets 1 virtual win and 1 virtual loss against an average paddler rated 1000, which keeps a first race from swinging wildly and anchors the scale. Every real race outweighs it quickly.

Rating = 1000 + 174 × log-strength (174 is 400 / ln 10, the chess scale). Two ratings give head-to-head odds: P(A beats B) = 1 / (1 + 10^((B − A) / 400)). A 100-point gap is 64 %, 200 points 76 %, 400 points 91 %. The ± band is one standard error, from the curvature of the fit.

What Moved The Rating

Each athlete page lists every counted race with the points it moved the rating. We start from 1000 and add the races in date order, each time finding the rating that best fits the races so far, with everyone else held at their current rating. The difference each race makes is its contribution, and the contributions add up exactly to the rating's distance from 1000. Whole numbers are rounded so they still add up.

Beside each race the page also prints its gap score: how far behind the top three you finished, in percent. It is the one number you can check against the results sheet with a calculator. It is not the rating.

Pools And Bridges

A rating is only comparable across regions when paddlers race across them. We sort races into pools (Europe, US mainland and Canada, Hawaii and Pacific, Australia and New Zealand, Asia and Middle East, Rest of the world) and count the bridge results in each: results by paddlers who also raced in another pool in the window. A pool with fewer than 20 bridge results is marked regional: its ratings compare within the pool only.

Release Checks And Fingerprints

Each list records a fingerprint of every result it counted, the method version and the constants, so it can be recomputed and checked: the same results always give the same list. Checks run on every list and tell us when something needs a second look: a new paddler in a top 20 on one race, a rating that moves more than 60 points, a merged identity in a top 50, a results file where more than 30 % of names matched no known paddler, or fewer results than the last list. A check never holds a list back.

Version History

Sources Of The Method

• Ralph A. Bradley and Milton E. Terry (1952), "Rank Analysis of Incomplete Block Designs: I. The Method of Paired Comparisons", Biometrika 39(3/4): the pairwise model.
• David R. Hunter (2004), "MM algorithms for generalized Bradley-Terry models", Annals of Statistics 32(1): the fitting algorithm.
• The Elo scale as used by FIDE, the world chess federation: FIDE Rating Regulations.
• The gap score follows the idea of FIS points in cross-country skiing: FIS Points Rules 2024/2025.
• Why bridges between regions matter: Massey's connectivity condition for ratings, arXiv 1701.03363.
• The method was tested before launch on public results: ICF Stand Up Paddling World Championships results (Long Distance and Technical, 2023 to 2025) and the SUP 11 City Tour 2026 overall results. Trained on earlier races, the ratings ordered later races as well as the other methods we tried, within a point or two, and their stated odds came true about as often as they said.

Back to the Power Score rankings →