The verdict
Most marine apps hand you ten dashboards. Submarius runs the integration step itself and gives you one of three outputs per dive day:
- GO — conditions favourable for the activity at the chosen location and time
- CAUTION — conditions workable but with notable risks or uncertainty; the breakdown explains what to watch
- NO — conditions unsuitable; the breakdown explains why
This is the verdict-first product philosophy: a verdict, not a dashboard.
How it’s computed
The verdict is activity-aware. The same conditions can be GO for fishing and NO for spearfishing, because the activities care about different signals:
| Activity | Dominant signals |
|---|---|
| Spearfishing / freediving | Water clarity, current, swell direction, surface chop, water temperature |
| Scuba diving | Surface conditions for entry/exit, current, water temperature, swell penetration vs site depth |
| Fishing (boat) | Wind, wave height, fronts, tide stage, bite-score window |
| Fishing (shore / wading) | Wind, surf, tide stage |
| Boating | Wind, wave height, fog, fronts, lightning risk |
For each activity, Submarius computes a per-signal traffic-light state (green / yellow / red) and combines them into the overall verdict using hand-tuned activity-specific rules — not a single weighted-sum score that treats a 30-knot wind the same as a missing tide table.
Why one verdict instead of ten numbers
Two reasons:
- Decision quality. When the user has to integrate ten numbers themselves, they integrate inconsistently and slowly, and they pick up on whichever number happens to be most salient that morning. A pre-integrated verdict is what an experienced captain would tell you if you called them.
- Accountability. A model that just shows ten numbers is never wrong. A model that says “GO” and then delivers two-foot visibility is wrong, and we can measure how often it’s wrong, calibrate against reports, and improve. Saying nothing is the easy way out.
When the verdict is suppressed
- Inland points. Submarius is a coastal product; we don’t compute verdicts for points far from any coastline.
- Insufficient data. If the satellite coverage is too stale, the forecast horizon too far out, or the location lacks a regional baseline, the app surfaces “insufficient data” rather than guess. Per the data-honesty principle: never claim what we don’t measure.
- Active hazards. Lightning risk above threshold, hurricane warning in effect, harmful-algal-bloom severity above 2 — these short-circuit to NO regardless of the rest of the forecast.
Drilling in
Tap the verdict to see the per-signal breakdown. Tap any signal to see the source data and the model’s reasoning. Nothing is hidden; every factor is reachable in three taps.