Surf Reports That Rate Themselves

Every spot gets a live rating on a six-step scale — Flat, Poor, Fair, Fair-Good, Good, and Epic — in the color language surfers already know: gray when there's nothing, blue and green as it builds, orange when it's good, purple when it's an event. Since Rating v4 (October 2026) it works the way a surfer reads a forecast — wind first, then size and period together, then how big this particular break can handle:

• Wind sets the ceiling. Every break carries its shore orientation, and wind is sorted into five classes by its angle to straight offshore: offshore, side-offshore, cross-shore, side-onshore and onshore. Each class has a ceiling that rises and falls with wind speed. Side-onshore over about 12 mph, cross-shore over about 18 mph, or onshore over about 10 mph rate Poor however big the swell is.
• Size and period are judged together. A knee-high 15-second swell is still knee-high; a big short-period wind swell is still chop. Short-period wind swell in real wind is capped at Poor.
• The bar rises with the break. The face needed for Fair-Good, Good and Epic scales with how big each break can hold. A clean five-foot day is a big day at Waikiki and an ordinary one at Pipeline.
• Good is rare, and Epic is an event. Good needs clean wind, a long period, and surf near the top of what that break handles. Epic needs all of that plus glassy or clean offshore wind.

A worked example — Cocoa Beach, Florida, October 17, 2026: 8 ft of 9-second swell under a 25 mph north-northeast wind. Under v3 that rated Fair to Good, because the wind there is side-onshore and v3 treated it as cross-shore. Under v4 it is Poor, which is what everyone standing on the beach would have called it.

The data is NOAA's GFS-Wave model sampled at each break's coordinates, refreshed with every model run — four times a day, out to 16 days.

v1 (June 21, 2026) shipped as a simple weighted score: period up to 45 points, size up to 30, wind up to 25, Epic at 80. The flaw took two weeks to show itself: period and size alone could reach 75 of the 80 points, and cross-shore wind cost nothing. On an ordinary July day, v1 rated 25 of our spots Epic — a third of the planet wearing purple, while Surfline called the same day Fair. Surfers know Epic is an event, and our scale was calling a normal Southern-Hemisphere week the day of the decade.

v2 (July 3, 2026) makes Epic an event: the wind gates above, a size-and-period floor for purple, and a rebalanced curve across every tier. Recalibrated against live data before shipping — the same day that produced 25 Epics under v1 produces 3 under v2 (Lagundri Bay, Macaronis, and Puerto Escondido — all long-period, chest-high-plus, and clean), with the rest of the world distributed the way a normal day should be: mostly Fair, a healthy band of Fair-Good, Good meaning genuinely good.

Every future calibration is versioned and ships as its own entry in the changelog — we tune this dial in public.

v4 started with a complaint from the beach. On October 17, 2026, a short-period wind swell with 25 mph north-northeast wind hit Florida's Space Coast, and v3 rated New Smyrna, Sebastian Inlet and Cocoa Beach Fair to Good. It was Poor.

So we checked the whole scale, not just Florida. We compared v3 with a leading commercial surf forecast at 75 of our breaks over four 16-day windows in January, April, July and September 2026: 56,255 daylight hours. We tuned v4 on half the breaks and scored it only on the other half, so it couldn't simply memorize them.

Surf heights changed too. The same comparison showed our face estimates running about 1.7 times the reference at a typical break. Each break now carries a measured height scale, so a spot page reads the way surfers compare surf. The three big-wave breaks the reference doesn't rate on the same scale (Waimea, Pe'ahi and Mavericks) keep their v3 heights.

Honest limits: Good is still our weakest call. We call it more often than the reference does and catch only about one in ten of its Good hours, because Good depends on details a swell model can't see. We keep collecting comparisons and will tune it in the open.

Here's the problem v3 solves. On July 3, 2026, the model showed Teahupo'o and Montañita, Ecuador nearly identical offshore swell — 3.1 ft at about 15 seconds. But Teahupo'o's reef rises out of deep water so abruptly that it focuses that energy into double-overhead slabs, while Montañita's gentler point turns it into head-high walls. Raw offshore numbers can't tell you that. Bathymetry can.

v3 estimates the breaking face height at each break in three steps, all published science:

• A physics baseline — the Komar & Gaughan (1972) breaking-wave transform, which grows waves as they shoal: long-period swell breaks meaningfully bigger than the same offshore height of wind chop.
• Each break's swell window — the compass arc of swell directions it actually receives. Swell aimed outside the window is attenuated, the way headlands and islands do it.
• A calibrated per-break factor — how this specific bathymetry amplifies or tames swell, by period band. The approach follows Caldwell & Aucan (2007), NOAA's Hawaii study that fitted exactly these coefficients against years of lifeguard observations (their reef fits reach ~2.1×). Ours are curated per break, published on each spot page, and bounded by that study.

The result — under v3, that same July day: Teahupo'o rates Epic at ~11-ft faces; Montañita rates Good at ~5-ft faces. Spot pages now lead with the face estimate ("Surf ~6–8 ft faces") and keep the raw offshore swell visible underneath.

Honest limits: ±20% is the physics floor; sandbars move, so beach calibrations are climatology, not this morning's bank; and island-shadowed coasts are the hardest problem in surf forecasting — ours included. When we tune a spot's calibration, it ships as a changelog entry.

Observed ratings (Forecast tab, Aug 2026): when a wave buoy sits within 50 km of a break and is either its forecast reference, upstream in its swell window, or the buoy the spot is mapped to, we run the buoy's measured swell — height, period and direction (spectral partition when available, else dominant period and mean direction) — through the very same v3 chain and label the result OBSERVED. Wind comes from the buoy only when it has an anemometer and sits within 30 km; otherwise from the model, and the badge says so. Buoys behind land or farther out show their reading without a rating, because an open-ocean number is not what arrives at the sand.