How Accurate Is Quiver's Surf Forecast?
Updated daily from live buoy data. Rolling 14-day window comparing our Quiver's forecasts against the NOAA marine baseline.
NOAA Baseline vs. Quiver
Quiver reduces wave height error from 0.679m to 0.614m — measured as Mean Absolute Error (MAE) against live buoy readings.
Lower is better. Values are average MAE across all tracked beaches (14-day rolling window).
Accuracy by Region
Wave height MAE (meters) by state — lower is better. Regions sorted by improvement over NOAA.
Only regions with 3+ validated beaches shown. 14-day rolling evaluation window.
Top Beaches by Accuracy Improvement
Ranked by reduction in wave height error versus the NOAA baseline. Minimum 20 validated predictions.
| # | Beach | NOAA Error | Quiver Error | Improvement |
|---|---|---|---|---|
| 1 | HB CliffsHuntington Beach, CA | 0.305m | 0.183m | +40% |
| 2 | Ocean Beach SF – SloatSan Francisco, CA | 0.605m | 0.424m | +30% |
| 3 | Ocean Beach SF – MiddleSan Francisco, CA | 0.594m | 0.442m | +26% |
| 4 | Manasquan InletManasquan, NJ | 0.677m | 0.523m | +23% |
| 5 | JobosIsabela, PR | 0.803m | 0.626m | +22% |
Help Us Get Even More Accurate
Every session you log helps calibrate our model. Report what you actually see at the beach — wave size, conditions, crowd — and your observations feed back into better forecasts for everyone.
Sign Up to Report ConditionsHow We Measure Forecast Accuracy
NOAA Marine Forecasts as the Baseline
NOAA's National Weather Service publishes marine wave height forecasts for coastal regions across the US. These serve as our baseline — they represent what any surfer would see checking a government forecast. Our ML model starts from these raw NOAA predictions and refines them.
Per-Beach ML Bias Correction
Every beach has unique exposure to swell direction, wind patterns, and local geography. Quiver's XGBoost model learns per-beach correction factors from historical buoy observations, adjusting NOAA's regional forecast to the specific conditions at each spot. Corrections account for terrain-aware factors like swell access and wind shelter across 72 directional bins.
Ground Truth from the IOOS Buoy Network
Validation uses real wave height readings from the Integrated Ocean Observing System (IOOS) buoy network. Buoys report significant wave height every 1–6 hours. We match each ML-corrected forecast to the nearest buoy observation within a 3-hour window to produce a verified prediction pair.
What MAE Means
Mean Absolute Error (MAE) measures the average absolute difference between a forecast and the observed buoy reading, expressed in meters. A raw MAE of 0.40m means the forecast is, on average, 0.4 meters off from what the buoy actually recorded. Lower MAE = more accurate forecast.
Rolling 14-Day Evaluation Window
All accuracy statistics shown here use a 14-day rolling window of matched forecast-observation pairs. This keeps the data current and reflects recent model performance rather than long-term historical averages. The data refreshes daily as new buoy observations become available.
Automated Weekly Retraining
The ML model retrains weekly using a 90-day rolling window of matched observation data. Retraining includes validation gates — the new model must outperform the previous version before deployment. This ensures the model stays calibrated as swell patterns and buoy data evolve over time.
Frequently Asked Questions
About surfing in surf forecast accuracy
- We use Mean Absolute Error (MAE) — the average absolute difference between a predicted wave height and the actual wave height recorded by a nearby IOOS buoy. Lower MAE means more accurate forecasts. We report MAE in meters and compare Quiver's ML-corrected forecast against the raw NOAA marine forecast baseline.
- Accuracy data updates daily. We use a rolling 14-day window, so today's numbers reflect the past two weeks of forecast-versus-buoy comparisons. This keeps the metrics current and avoids overfitting to seasonal averages.
- Ground truth comes from the Integrated Ocean Observing System (IOOS) buoy network — a national network of ocean sensors operated by NOAA. Buoys report significant wave height every 1–6 hours. We match each forecast to the nearest buoy observation within a 3-hour window.
- The NOAA baseline is the raw National Weather Service (NWS) marine wave height forecast — unmodified. It represents the regional average forecast that any surfer can access from weather.gov. Quiver's ML model starts from this baseline and applies per-beach corrections. The improvement percentage shown is how much smaller our error is compared to using the raw NOAA forecast.
- NOAA publishes regional forecasts that don't account for individual beach geography. Quiver's XGBoost model learns per-beach bias correction factors from historical buoy observations. It adjusts for each beach's unique exposure to swell direction, wind patterns, and terrain features like offshore reefs and headlands. The model retrains weekly on a 90-day rolling window of matched data.
- We currently track forecast accuracy for 79 beaches with at least 10 validated forecast-buoy pairs. Beaches need sufficient matched data to appear in accuracy statistics. As more buoy observations accumulate, additional beaches are added automatically. New beaches typically reach minimum data thresholds within 2–4 weeks of activation.
Ready to check the forecast?
Get ML-corrected surf forecasts, tide charts, and crowd levels for beaches across the US.