Effort tiers
The search works forward in time, asking at each step where the boat could reach from everywhere it could already be. How finely it asks is not a setting. The heading fan, the time step, and the spacing at which nearby positions are considered distinct are derived from your course and from the forecast itself — the grid spacing and time cadence of the data being searched. Resolving finer than the data would sharpen the interpolation between grid points, not the weather; resolving finer than your course demands (a mark rounding's approach angle, a channel's turns) is owed to every route, so every tier gets it. Courses with marks get a 1° heading fan; open water gets 2°, where the exact tacking and gybing angles are always tried regardless.
What a tier buys is the forecast: more steps, faster cadence, and models decoded on finer native grids — and with better data, the derived search sharpens to match automatically.
| Tier | Forecast steps | Model grids | Returns in | Plan |
|---|---|---|---|---|
| Standard | Baseline set | to 0.25° | seconds | Included |
| All-out | Every published step, 15-minute HRRR | ~5 km native | minutes | Pro |
There is nothing to pick. Every compute runs at the strongest tier your plan includes: signed out or on a free account, that is the full derived search over the baseline forecast set — Standard, whose answer on that data is the answer, with no finer search to buy. On Pro, every compute is All-out.
All-out takes everything the data supports: every published step of every model out to the forecast's own horizon, HRRR near its native ~5 km grid with 15-minute steps instead of resampling to the corridor's 0.25°, and NAM the same way out to its 60-hour horizon. Nothing is thinned, ever — what the model publishes is what the search samples. Where the mesoscale models cover you and the weather is changing quickly, this is the tier that sees it.
The time window is automatic too
There is no "time limit" field. The forecast window is sized from the course — a nominal passage
time with a weather margin, never less than a full day, capped by the forecast's own horizon. A
passage longer than the forecast is still accepted and returns honestly as horizon where the
data ends.
Changing your mind is cheap
Computing again at a different departure, course or wind scale re-uses the forecast already downloaded, returns quickly, and spends nothing from your allowance.
Long passages
On a multi-week passage the fully derived search can exceed what one machine should spend or hold. When that happens the dials are walked back — a wider fan, a longer step, a narrower frontier — and the result says so in its warnings. The forecast is still ingested at the tier's full rate; only the search coarsens, never the data.