BestRacesMethodology · last computed October 30, 2026
How the rankings work
Most "best races" lists don't tell you how they were made. This page is the entire recipe. A ranked race carries a composite score out of 100, computed from four components using public signals. The same formula applies to every race — we don't hand-adjust positions, and no one can pay for one.
It also tells you what we haven't measured. Of the 6,984 races we list, 683 currently carry a score. The other 6,301 are listed and searchable but deliberately unranked — read on for why.
The four components
| Component | What it measures | Weight |
|---|---|---|
| Field size | Participation, on a log scale so strong regional races aren't drowned out by the majors. Trail races are scored on their own curve — a 400-runner ultra is "big." | 30% |
| History | Years since founding, credited up to 50 years. Longevity is the hardest signal to fake. | 20% |
| Credentials | USATF course certification, Boston-qualifier eligibility, World Marathon Major status. Trail races receive a baseline here since certification doesn't apply to them, and multisport races read USA Triathlon sanctioning instead. | 25% |
| Character | Course scenery (editorially assigned, 1–5, defined below) and, for trail races, terrain drama measured by elevation gain. | 25% |
How scenery is rated
Scenery is the one number in the formula a person sets, so here is exactly what it means. We rate what a runner sees from the course on race day: water, mountains and open country, parks and forest, landmarks and historic streets, skyline views, and seasonal scenery a race is timed for. Crowds, costumes, swag and the cause don't count. They matter, but they aren't scenery.
| Rating | What it means | For example |
|---|---|---|
| 5 | Destination scenery. Most of the course is on or beside something people travel to see: coastline, mountains, a national park, an iconic skyline or monuments. | Big Sur, Cherry Blossom Ten Mile |
| 4 | Consistently scenic. Well over half the course runs along water, parkland, historic districts or landmark streets. | Chicago, Eugene |
| 3 | Pleasant, with highlights. A mix of park, river or downtown sections and ordinary roads. | Houston, Broad Street Run |
| 2 | Mostly ordinary. Suburban roads, neighborhoods or business parks, with a short scenic stretch at most. | |
| 1 | Unappealing. Mostly frontage roads, industrial areas or parking lots. |
We rate the route itself, from the course map and imagery along it, not from race photos, which are chosen to flatter, and we use OpenStreetMap to measure how much of the route runs past water, parks and landmarks. Ratings are by request: the director of a claimed race can ask for one, and every request is rated on this scale in batches. A rating can raise or lower a score, and it stands either way. Nothing a race does with us, paid or free, changes it. 39 courses carry a rating so far. Until a course is rated, scenery counts as unmeasured, which scores the same as a rating of 2.5 (see below).
What an unmeasured signal is worth
Nothing. It is left out of the arithmetic entirely, and the remaining weights are renormalized over what we did measure — so a race scored on field size and history is scored out of 100 on those two, not out of 50 with the rest counted as zero. The distinction matters: scoring an unknown as zero would assert that a race has no history, when the truth is only that we haven't checked.
Renormalizing alone would reward missing data — a race measured on two strong components would score as if the other two were just as strong. So the renormalized score is pulled toward a neutral 50 in proportion to the weight we did not measure: a race scored on half the formula tops out at 75, and passes a fully-measured race only by being much stronger on what we did measure.
Every race page shows which components were measured and which weren't, so you can see how much of the formula a given score actually rests on.
When a race gets ranked
When at least 2 of the four components are measured and they account for at least 45% of the total weight. Below that bar a race is listed without a score — no number, no position, no bib. One measured component renormalized to 100% of the weight looks identical to a fully-measured score and isn't, so we don't publish it.
A race below that bar can still be ranked on the evidence we do have. 396 of them tell us their own age in their name — a "38th Annual" is the race stating how many times it has been run — so they appear in a longest-running list, ordered oldest first, on every page that covers them. That list shows founding years rather than scores and is never labeled "best": ordering races by age is a claim the evidence supports outright, and calling the oldest race the best one is not.
This is why most of our catalog is unranked today. We sync thousands of races from public registration data, which reliably gives us name, place, date and distance — and almost never gives us field size or founding year. Here is where we stand on each signal:
| Component | Races measured | Share of catalog |
|---|---|---|
| Field size | 2,836 | 41% |
| History | 888 | 13% |
| Credentials | 466 | 7% |
| Character | 39 | 1% |
Those numbers are generated by the ranking engine on every build, not typed in here, and we expect them to climb. Until they do, we would rather show you a gap than invent a number to fill it.
Questions we get asked
How are races scored?
A ranked race carries a composite score out of 100, computed from four weighted components: field size (30%), history (20%), credentials (25%), and character (25%). The same formula runs over every race in the database — we don't hand-adjust positions, and no one can pay for one. Components we haven't measured are left out of the arithmetic rather than counted as zero, and the remaining weights are renormalized over what is known — then the result is pulled toward a neutral 50 in proportion to the weight we did not measure, so a half-measured race cannot outscore a fully-measured one on missing evidence.
Why do some races have no score?
Because we haven't measured enough of them to rank them honestly. A race earns a score once at least 2 of the four components are measured and they cover at least 45% of the total weight. 683 of the 6,984 races we list currently clear that bar. Another 396 state their own age in their name, so they are ranked by how long they have been run — oldest first, labeled as a longest-running list rather than a best-of, and showing founding years instead of scores. The rest stay listed and searchable with no score, no position and no rank number. Public registration data reliably gives us a race's name, place, date and distance, and almost never gives us field size or founding year — so we show the gap instead of inventing a number to fill it.
Can a race pay for a better rank?
No. Rankings are never sold. Paid products do exist — featured listings and brand sponsorships — but they are visually distinct, labeled as paid, render outside the ranked list, and can never change a race's composite score or its position. A database-level guard blocks payments, claims, and director edits from writing to any ranking score, so the rule is enforced in the code rather than only promised here.
Do affiliate links affect the rankings?
No. Some registration links on this site are affiliate links — if you register after clicking one, the registration platform shares a small portion of its processing fee with us at no extra cost to you. The scoring formula has no input for which platform a race registers on, so affiliate status cannot move a race up or down.
What is a featured partner race?
A race whose director has claimed it here can become a featured partner. The race gives us a referral code or link from its own registration platform and pays us a referral fee when someone registers through it: 15% of the entry fee, at least $3. In return, when a search on the map or the search page finds races near it, the race can appear in one card marked "Featured partner" above the results. The card shows no score, and the race still appears in its normal place in the list. Being a partner never changes a race's score, its rank, or which lists it is in. It costs runners nothing extra: the race pays the fee, and the entry price is the same.
How is field size measured?
Field size is participation in the race's headline distance — not the combined total across every distance offered that weekend — scored on a log scale so strong regional races aren't drowned out by the majors. At 30% it is the heaviest single component.
How is race history weighted?
History is 20% of the score: years since the race was founded, credited up to 50 years. Longevity is the hardest signal to fake — a race in its fortieth year has survived four decades of weather, permits, and volunteer turnover.
What counts as a credential?
Credentials are 25% of the score and cover USATF course certification, Boston Marathon qualifier eligibility, and World Marathon Major status — objective marks awarded by outside bodies, not anything we assign.
What is the character score?
Character is 25% of the composite: course scenery, editorially assigned on a 1–5 scale, plus terrain drama measured by elevation gain for trail races. It is the one component that involves human judgment, which is why each level of the scale is defined above rather than folded quietly into the total.
How are triathlons and duathlons scored?
Swim-bike-run events are ranked in their own category, never against running races, and on their own scale: field size is judged against other multisport races, where a few hundred finishers is a big field, and the credential component reads USA Triathlon sanctioning instead of USATF certification — USATF measures running courses, so a triathlon can never hold one. 19 of them have their sanctioning on record.
Why are trail races scored differently?
Trail races are scored on their own curve. Field size is judged against other trail races — a 400-runner ultra counts as big — and because road-course certification doesn't apply to them, they receive a baseline credential score instead of being penalized for a certificate that cannot exist. Their character score also picks up elevation gain.
How often are the rankings recomputed?
The entire site is rebuilt daily and every score is recomputed from scratch in that build — no ranking is cached or carried forward by hand. The scores shown here were last computed on October 30, 2026.
Where does the race data come from, and how do I correct an error?
Race data comes from registration-platform APIs, public race calendars, and manual research. Race directors can claim their listing for free to correct dates, distances, and details; anyone else who spots an error can tell us and we'll fix it. A claim changes the facts on a page, never the score the formula computes from them.
What we will never do
Rankings are never sold. Paid placement is visually distinct from earned rankings, clearly labeled, and never affects a race's composite score or position. That commitment was written here before any paid product existed, so you can hold us to it.