Methodology
How accurate is PickADive?
No dive-conditions forecast is a guarantee — not ours, not anyone’s. What a conditions page can do honestly is show you exactly how each number was built, what kind of data it came from, how old it is, and how far away it was measured or modeled. This page explains how PickADive does that, and how to evaluate any conditions forecast you rely on, including ours.
Last updated: July 23, 2026
The four kinds of dive conditions data
Every number on a conditions page is one of four kinds of data, and they are not equally reliable for the same question. PickADive labels each kind distinctly so you always know which one you are reading.
| Data kind | What it is | How PickADive labels it |
|---|---|---|
| Forecast | A weather or marine model’s best prediction of conditions that have not happened yet. | Shown as a “Model estimate” or “Forecast” with a timestamp for when the data was gathered. |
| Measured observation | An instrument reading of conditions that actually occurred at or near the site, such as recorded water temperature. | Labeled as a measured observation in the conditions evidence, separate from model predictions. |
| Historical typical | What conditions usually look like for that place and season — an average over past years, not a prediction for a specific day. | Presented with “typical” wording, such as a site’s typical best season, so it cannot be mistaken for this week’s outlook. |
| Diver report | A firsthand account from a person who was at the site and shared what they experienced. | Shown as “Diver report” with the dive date, and given priority over model estimates while it is recent. |
For how to read these estimates on an individual site page — windows, details, and limited-context states — see the dive conditions guide.
What PickADive shows that most conditions apps don’t
Accuracy starts with showing your work. These details appear on PickADive site pages precisely because they let you second-guess us.
- Model-point distance: marine models compute values at grid points, not at your entry rocks. When an estimate comes from a distant point, the site page says so — for example, “Model estimate from a point 4 km offshore.” Across sites with marine data, that distance currently averages about 6.3 km.
- Named sources: forecast data comes from Open-Meteo and NOAA, and both are credited on the pages that use them — not buried in fine print.
- Timestamps on every value: estimates show when the underlying data was gathered or when a window was computed, so stale context is visible instead of silently presented as current.
- Honest empty states: when a provider has no useful signal for a site, the page says “no data” instead of inventing a number. A metric with nothing behind it simply does not render.
The quality standards behind the dive-site catalog itself — what gets published and what stays out — are covered on the dive-site data quality page.
A diver’s checklist for judging any conditions forecast
These five checks apply to every conditions app or forecast page, including this one. If a source fails several of them, treat its numbers as decoration rather than data.
- Does it show its sources? A trustworthy conditions page names where its numbers come from. If you cannot find a data source anywhere on the page, you cannot judge the number it produced.
- Does it explain what drives the number? A single score or icon with no visible inputs is a black box. Look for the underlying wave, wind, temperature, or report details — and an explanation of how they combine.
- Does it show how far away its data comes from? Marine models compute values at grid points that can sit well offshore from the spot you plan to enter the water. An estimate from a distant point can differ meaningfully from conditions at the site itself, so the distance is part of the answer.
- Does it admit gaps? No provider has useful signal for every site on Earth. When data is missing, an honest page says so. A page that always shows a number for every site is filling gaps with something.
- Can users correct it? Local divers usually know a site better than any model. A conditions page that accepts reports and corrections gets better over time; one that cannot be corrected repeats its mistakes.
Why there is no accuracy percentage on this page
A claim like “95% accurate” needs a rigorous, published comparison between forecasts and what divers actually found in the water — across sites, seasons, and conditions. PickADive does not yet have that evidence base, so it does not publish a score. As diver reports accumulate, they become the measuring stick the estimates are checked against. Until then, this page explains the method, not a marketing number.
How to check us
The fastest way to test a conditions page is to compare it with what you found in the water. PickADive is built to absorb that feedback.
- Report conditions after a dive: every site page accepts a short report of what you actually experienced. Recent diver reports are shown alongside — and take priority over — model estimates.
- Suggest a correction: if a depth, access note, hazard, or map position is wrong, you can flag it directly from the site page. The most useful corrections say what is wrong and what local divers would recognize as right.
- Evidence-based verification: site facts are backed by a verification process that ties each claim to its supporting evidence, with public labels — Verified, Diver reported, Needs update — rolling out across site pages.
Corrections and reports are reviewed before they change what other divers see, so one submission cannot silently overwrite a page. For the broader standards behind this process, see the PickADive Trust Center.
Use it as planning context, nothing more
All conditions information on PickADive is forecast-only planning context — always assess conditions on site and dive within your training and limits. Local operators, current observations, your equipment, and your judgment matter more than any estimate on any page.
Ready to put the checklist to work? Open any site from the dive site map and look for the sources, distances, timestamps, and gaps yourself.