False-alarm rate requires at least one resolved negative outcome.
sounding open dataReading the waterlineLive sources are loading. Nothing is filled with placeholder market data.
forecast quality, without selective wins
Bathymark publishes calibration only when the Evidence Ledger contains comparable, precommitted binary forecasts and sourced outcomes. Event pressure without a probability is not backfilled into a probability claim later.
Pressure windows can remain useful as sourced event context without pretending to be probabilistic forecasts. The complete denominator stays visible while calibration is still building.
Brier score requires a precommitted probability between 0 and 1 and an explicit resolved binary outcome.
0 eligible binary records across 52 published windowsA headline win rate cannot replace coverage, false alarms, recall, or probability calibration.
False-alarm rate requires at least one resolved negative outcome.
Precision requires at least one explicit positive classification.
Recall requires at least one resolved positive outcome.
Brier score requires a precommitted probability between 0 and 1 and an explicit resolved binary outcome.
Classification eligibility requires a binary target, prediction, threshold, horizon, and sourced outcome. Brier scoring additionally requires a probability fixed before resolution. Current Tide Map windows intentionally publish no probability when the model cannot support one.