Catch likelihood
The primary outcome label we intend to calibrate for a planned effort and duration. It is not currently calibrated in this preview.
CastingCompassABOUT CASTINGCOMPASS
Useful context before the cast.
CastingCompass turns public coastal information and your planned fishing context into explainable guidance. The product helps you choose where and when to look; it does not promise a catch.
READ THE LABELS
Five ideas, kept separate.
These labels answer different questions. Keeping them distinct is how we avoid turning a comparative rank into false precision.
Expected catch rate
A secondary metric for catches per angler-hour when there is enough reviewed effort-normalized trip data.
Opportunity Rank
A comparative 0–100 ordering of the options currently evaluated. It is not catch likelihood and is not a probability.
Fishability
A practical view of access, conditions, legal or posted constraints, safety context, and whether an angler can work the water effectively. It does not establish water safety.
Data Confidence
A signal about freshness, source support, and extrapolation. Sparse or stale evidence widens uncertainty rather than becoming false precision.
WHAT INFORMS THE PLAN
A layered view of the water.
Every layer has boundaries. Editorial location guidance stays distinct from learned or statistical model output, and missing evidence is treated as uncertainty.
01Location, habitat, and bathymetry
Curated access metadata, shoreline structure, habitat tags, and charted bathymetry describe what is known about the place. These are context and evidence gaps, not a guarantee that fish are present.
02Seasonality and conditions
Seasonality, tide and current, wind, swell, water temperature, daylight, moon context, and source freshness shape the planning view. Stale or missing inputs stay visible as uncertainty instead of silently becoming current.
03Planned effort and fishing method
A responsible catch estimate needs a planned duration, access mode, and fishing method. The current staging preview uses two-hour windows for comparison, but it does not yet present a calibrated catch likelihood.
04Trip-log feedback
Completed trips add structured effort and outcome context. Catches and skunks are useful feedback, but they do not automatically train or validate the live ranker. Future calibration will use reviewed, effort-normalized records.
MODEL PATH
Transparent first. Calibrated next.
Explainable baseline
The current planner is expert-configured and untrained. It combines reviewed static context, seasonality, current conditions, practical fishability, and source freshness into comparative planning guidance.
Effort-normalized calibration
The future path is a calibrated model with location pooling: planned duration, method, access mode, conditions, and trip outcomes can support catch likelihood and expected catch rate without requiring deep learning by default.
Important: the future path is not already deployed. Scores are estimates, not guarantees. Sparse data, stale sources, and extrapolation widen uncertainty.
AUTOMATION & PRIVACY
Bounded tools, clear limits.
Some features may use automated assistance, but the product should expose useful structure rather than hide decisions behind a black box.
Fish-photo analysis
When available, photo analysis is bounded to structured results such as a possible species label or review flag. It is not a definitive identification, legal determination, or substitute for local regulations.
Data provenance
Public sources, review dates, model versions, and source freshness are part of the product context. Provider credentials and private implementation details are never exposed in the UI.
Human judgment
Confirm current regulations, access restrictions, closures, weather, water conditions, and your own safety plan before fishing. CastingCompass cannot see every change at the shoreline.