Forecast accuracy

How wave-height forecasts are checked, where buoy observations help, and where they do not.

Start with the measurement

How to judge a surf forecast.

A useful accuracy claim names the forecast variable, saves the prediction before conditions arrive, and checks it against the matching observation. If those pieces differ, the score does not support a ranking.

Quiver's current position

No same-sample comparison. No accuracy ranking.

Quiver has not evaluated other surf forecasts on the same beaches, timestamps, lead times, and wave-height definitions. This page explains the method we expect any comparison to meet.

Forecast
Save the value, place, valid time, lead time, and units.
Observation
Match it to a later reading from a defensible ground-truth source.
Definition
Confirm both values describe the same physical quantity.

Same words, different water

Offshore wave height is not breaking face height.

A buoy and a surfer can both report “wave height” while measuring different physical quantities. A valid score has to keep that distinction visible.

Buoy observation

Significant wave height offshore

IOOS and NOAA buoys estimate significant wave height, often written as Hs. It describes the offshore sea state around the sensor, not the face of a wave breaking at a beach.

Surf observation

Breaking face height at the beach

The wave a surfer sees depends on bathymetry, exposure, refraction, tide, and where it breaks. It cannot be scored directly against offshore Hs without a documented conversion and on-beach observations.

A buoy check can validate an offshore-height forecast. By itself, it cannot prove the breaking surf-height call was right.

See the method →

A fair test

Measure the same thing on the same sample.

Mean absolute error can summarize the average miss, but the number is only comparable when the inputs follow the same rules.

  1. 01 save

    Freeze the forecast

    Store the forecast before the observation arrives, including its issue time, valid time, location, lead time, variable, and units. Do not backfill a revised value.

  2. 02 match

    Pair the observation

    Match the saved value to the nearest defensible observation in a fixed time window. For automated wave-height checks, Quiver uses time-matched readings from nearby IOOS and NOAA buoys.

  3. 03 report

    Publish the limits

    Report sample size, beaches, date range, forecast horizon, missing data, and the exact error formula. Separate offshore Hs checks from breaking-wave observations.

What a head-to-head requires

Every forecast must cover the same beaches, valid times, lead times, observations, wave-height definition, and exclusion rules. Different samples can describe different conditions and reverse the apparent result.

What Quiver can claim today

Quiver has not completed that same-sample comparison against Surfline, NOAA, or another surf forecast. We therefore do not claim that Quiver ranks first on accuracy.

Your report is a field note, not ground truth.

Log what you saw: wave size, conditions, and crowd. Reports are captured for operator review. They do not automatically train or calibrate production forecasts.

Log a session

Straight answers

Forecast accuracy FAQ

Frequently Asked Questions

About surfing in surf forecast accuracy

For automated wave-height checks, Quiver pairs a saved forecast with a time-matched observation from a nearby IOOS or NOAA buoy. That observation is offshore significant wave height. The station, time window, forecast horizon, and missing-data rules must travel with any reported error score.
Significant wave height, often written as Hs, is a statistical description of the sea state around a buoy. It is not a direct measurement of the face of a wave breaking at a beach.
As swell reaches shore, exposure, bathymetry, refraction, tide, and the breaking point change the wave a surfer sees. A face-height forecast needs on-beach observations or a documented conversion before it can be validated against offshore Hs.
Quiver does not claim that ranking. A same-sample comparison has not been completed across identical beaches, timestamps, lead times, observations, and wave-height definitions.
Freeze every provider's forecast before conditions arrive, evaluate the same variable at the same places and times, use the same ground truth and exclusion rules, then publish the sample size, date range, forecast horizon, and missing data.
No. Session notes and condition reports are captured as field feedback and reviewed by operators. They do not automatically train or calibrate production forecasts.
Check whether the forecast names the variable it predicts, saves predictions before the outcome, uses matching observations, separates offshore height from breaking surf height, and publishes misses as clearly as hits.