How scoring works

DontDieFishing turns the available launch-site conditions into a 0–100 planning score for a selected vessel. This page is the canonical explanation of the shared scorer: what it reads, how it normalizes each signal, how your vessel changes the inputs, and when a hard veto forces NO-GO.

The scoring pipeline

For each conditions snapshot, the scorer follows this order:

1. Select thresholds. It starts with the default thresholds for the vessel type, then applies any custom threshold overrides.
2. Select weights. It starts with the chosen cautious, moderate, or experienced profile, then applies any custom weight overrides.
3. Check hard vetoes. A veto is evaluated before weighted scoring. If one fires, the result is immediately 100, NO-GO, with a reason.
4. Build signal rows. Each signal records its raw value, normalized risk, availability, adjusted weight, and contribution.
5. Normalize available weights. Missing signals contribute no weight; the remaining available weights are redistributed proportionally.
6. Combine and classify. Contributions are summed from unrounded values, then the result is rounded, clamped to 0–100, and mapped to GO, CAUTION, or NO-GO.

Conditions and signals

The web aggregator combines these sources before scoring:

Tide height and phase, sunrise and sunset, air temperature, wave period, and swell direction can be shown as condition details. They are not weighted signal rows in the shared scorer.

Normalization

Every available signal is converted to a 0–100 risk value before weights are applied. Lower values represent lower modeled risk.

Threshold-based signals

Wind speed, gusts, wave height, and tidal current use a piecewise scale:

• 0 at the ideal value (0 for these signals)
• 70 at the selected vessel threshold
• 100 at 1.5 times that threshold

Values between those anchors are interpolated. Wind / gusts use the higher of the two normalized values. Waves / swell uses wave height in the scorer; period and direction are not additional normalized inputs.

Visibility and water temperature use an inverse scale:

• Visibility is 0 at 10 miles, 70 at the selected visibility threshold, and 100 at half that threshold.
• Water temperature is 0 at 70°F, 70 at the selected water-temperature threshold, and 100 at half that threshold.

Categorical and direct signals

Vessel personalization

The vessel type controls the default environmental thresholds used by the normalizers and veto checks. The defaults are:

The scorer does not read a vessel-specific lightning threshold. high lightning is a hard veto for every vessel; lower lightning levels remain weighted signals according to the normalization table.

Custom thresholds are applied after these defaults. A vessel also selects one of three weight profiles. The values below are the base weights before missing signals are removed and the remaining weights are renormalized:

Custom weights are applied after the profile defaults. The scorer then sums the weights of available signals and divides each available base weight by that sum. This keeps the available signals' adjusted weights proportional when a source is missing.

Hard vetoes

The scorer checks these conditions before it computes any weighted signal contribution:

• Any marine warning containing small craft advisory, hurricane, tropical storm, gale warning, storm warning, hurricane warning, or hurricane watch.
• Lightning risk equal to high.
• Wind speed greater than 1.5 times the vessel's wind threshold.
• Gusts greater than 1.5 times the vessel's gust threshold.
• Wave height greater than 1.5 times the vessel's wave threshold.
• Visibility less than half the vessel's visibility threshold.
• Tidal current greater than 1.5 times the vessel's current threshold.
• Water temperature below 75% of the vessel's water-temperature threshold.

A veto result has:

• total_score: 100
• status: "nogo" (shown as NO-GO)
• vetoed: true
• veto_reason containing the matching warning or measurement
• an empty signal list, because the veto returns before signal rows are built

A warning shown in the UI is not automatically a veto. The marine-warning callout displays any active string warning, notes that it may have changed, and explicitly directs you to check official NWS guidance. Only the warning phrases in the veto list force NO-GO.

Status bands

When no veto fires, the final rounded score is classified as follows:

A veto is NO-GO regardless of what a weighted score would have been. The score is not a guarantee that a trip is safe, and GO is not an authorization to launch.

Freshness, missing data, and degraded fetches

The web conditions endpoint reuses a cached snapshot only when it is less than 30 minutes old. Without a fresh snapshot it calls the aggregator. The NWS fetchers currently degrade independently when a request fails: the alerts fetcher returns an empty alert list, while the forecast fetcher returns lightning_risk: "none" and null values for the forecast measurements it could not obtain. The aggregator passes those fallback values to the scorer without an error marker.

That behavior is fail-open, not fail-closed. An empty alert list cannot trigger an alert veto, and the fallback lightning value maps to zero modeled risk. Unknown alerts or lightning can therefore contribute to a GO result. Other missing numeric signals are represented as unavailable; their normalized value is 0 and their adjusted weight is 0, with the base weight redistributed among available signals. A fetched_at timestamp does not prove that every source was successfully fetched, and returned signal rows currently have is_stale: false.

This is a product limitation under review. Treat a missing, unverified, or degraded input as unknown: verify current conditions directly with authoritative NWS and NDBC sources before launch, and make the launch decision with current conditions and your own judgment. Do not read an absent or degraded score as authorization to launch.