Methodology

How the Hunt Score is calculated.

The Hunt Score is a 0–100 forecast of how productive a tooth-hunting beach is likely to be on a given day. It is not a guarantee — it is a calibrated heuristic built from public, auditable data sources.

Inputs

NOAA CO-OPS — predicted tide extremes (high/low) and harmonic curves for the nearest tidal station.
Open-Meteo — hourly wind speed, gust, and bearing for the site coordinates.
SunCalc — sunrise, sunset, and lunar phase derived from the date and lat/lon.
NDBC buoys (where available) — significant wave height for storm exposure.
Site dossier — score model (tidal-beach, river-low-water, spoil-pile, inland-creek), good wind direction, and access type.

Components (tidal-beach model)

Each row below is an additive component of the daily score, with its exact maximum in points. The sum is then multiplied by a storm modifier (1.00–1.30×) and an event boost is added on top, then the result is clamped to 0–100. Every per-day score on a location page exposes the same numbers in the “Show the math” toggle so you can verify by hand.

Tide range

0–50 pts

Day’s max-high minus min-low. Linear ramp from 0 ft (0 pts) to 6 ft (50 pts), saturated above 6 ft.

Spring-tide proximity

0–15 pts

Exponential decay from the nearest new or full moon, half-life 3.5 days. Within ±2 days of syzygy you get nearly the full 15 pts.

Daylight window

0–15 pts

Smooth Gaussian curve centred on golden hour (~1.5h after sunrise). Daylight floor 0.4 × max so a noon low still scores well; night lows still get a small floor (1.5 pts) because some hunters use lights.

Recent stir

0–10 pts

Past-72h average onshore wind speed, projected onto the site’s good-wind direction. Only counts hours that actually happened — forecasts of past wind are ignored.

Wind setup at low

−10 to +10 pts

Forecast wind during the ±4h low-tide window. Offshore wind drains water (more beach, +pts); onshore piles it up (less beach, −pts). The vertical setup is approximated as ~0.04 ft per mph of sustained wind, clamped at ±2 ft.

Surf bonus

0–5 pts

Recent significant wave height. 1–4 ft = clean sorting (3–5 pts); >7 ft = unfishable (0 pts).

Recent rainfall

0–3 pts

Past-48h precipitation. Heavy rains flush creek and inlet material onto the strand. ≥0.1\" = +1, ≥0.25\" = +2, ≥0.5\" = +3. Smaller weighting than cliff-beach, where clay slumps drive output (max +12).

Storm multiplier

×1.00–1.30

Multiplicative on the component sum. Triggered by peak gust thresholds: >25 mph = 1.15×, >35 mph = 1.30×. Reflects the post-storm churn that uncovers buried material.

Event boost

+0 to +25

Hand-curated events plus auto-generated signals (drought, flood watch, dredge cycle). Capped at +25 and, for cliff-beach sites, scaled down on drowned-beach days so a manually curated +15 doesn’t override the wind telling us nobody can hunt.

Other site models

river-low-water — gauge level + recent rainfall instead of tide. Best when water is dropping and gravel bars emerge.
spoil-pile — dredge or quarry sites. Tide-independent. Penalizes high-wind dust days, slight bonus after heavy rain rinses fresh material.
inland-creek — small drainages where exposure depends almost entirely on recent storm flows. Bumps after rain, fades with drought.

What the colors mean

85+

Prime window

70–84

Strong

55–69

Workable

< 55

Off-day

Reliability tiers

Each site advertises how trustworthy its live score is:

full — NOAA tide station + Open-Meteo, all model inputs available.
partial — model runs but one input is approximated (e.g. nearest buoy > 50 km away).
dossier-only — global heritage / inland sites with no live oceanographic data. The dossier is real; the daily score is a baseline placeholder.

Not currently in the model

Transparency goes both ways — here is what we deliberately don’t use today, and why:

Hunt pressure (info-only). The number of hunts logged at a site in the past 7 days is shown as a chip on each location page, but does not change the score yet. We need more volume per location before a penalty would be statistically meaningful.
Wave period and swell direction. We use significant wave height only. Open-Meteo doesn’t expose period reliably for most sites.
Barometric pressure trend. A falling barometer is an early storm signal, but our existing wind-based stir already captures the same effect once the wind arrives.
Water temperature, clarity, salinity. Affect comfort and snorkel visibility, not (directly) what teeth wash up on the beach.
Weekday vs weekend. A 70 on Tuesday is often better in practice than a 70 on Saturday because of crowds, but we don’t adjust for that.
Beach renourishment. Fresh sand buries everything. Captured only when an admin logs a manual event — there’s no auto-detection from satellite imagery yet.

Open source & auditable

The full scoring implementation lives in src/lib/huntScore.ts in our public repository. If you spot an error or have a calibration suggestion, file an issue.

Calibration is ongoing. We compare predicted scores against logbook outcomes (when users opt in) and adjust weights quarterly.

See it live on the map About Toothhound

Last reviewed: April 2026