We publish these numbers nightly, automatically, whether they flatter us or not —
because accuracy is the product, not hand-picked screenshots. Everything below is
measured against something we did not control: buoys in the water, and the ocean
that actually showed up. Dotted terms have definitions — hover or tap them.
How to read this page
There are three different questions here, and they get harder going down:
1. Is today’s water temperature right? Easiest to check — buoys report
real temperatures, so we just compare.
2. Was yesterday’s forecast right? Harder — we score old forecasts against
what actually landed.
3. Did the model find fish? The one that matters, and the slowest to answer,
because it needs logged catches to accumulate.
1. Today — sea surface temperature vs. buoys
Our satellite SST at each NDBC buoy’s exact position, against what that buoy
measured in the water. A buoy is ground truth: when it disagrees with us, we are wrong.
2. Rolling window
The same comparison averaged over several days, so one cloudy morning
cannot flatter or embarrass us. This is the number to judge us on.
3. Forecast skill — how fast does the outlook decay?
We archive every forward-looking outlook, then score it against the analysis that
eventually landed for that same date. A forecast should get worse the further out it
reaches — if these numbers were flat, it would mean the forecast was not really
forecasting. Until these hold up over a full season, the app labels every
projection “outlook” and we make no forecast-accuracy claim.
Lead
Days scored
SST miss (°C)
Habitat drift
4. Did the model find fish?
The only question that ultimately matters, and the one no competitor can answer:
when captains log catches, do those catches land in the water the model called good?
We compare the habitat score at real catch locations against the average score across
the whole region that day — i.e. against what picking water at random would have given you.
Species
Catches scored
Avg score at catch
Random water
Score lift
Prime lift
5. Bathymetry source change — NOAA CRM
This is an input-change report: it shows how much the depth field, the
habitat score, and the recommendation pins moved when the bathymetry source changed.
It is not an accuracy measurement and makes no claim that predictions improved.
Validating that requires survey truth or catch-conditioned scoring, neither of which
exists yet.
Water depth
Cells compared
Median change (m)
99th percentile (m)
Species
Prime before
Prime after
Cells reclassified
Daily breakdown
Every day in the rolling window, unaggregated, so you can see the spread rather
than trust an average.