Actual POD interface with invented records, forecasts, scores and timings. These examples are not measured results.
Compare models, check calibration and inspect the distributions behind their scores.
Compare the same historical cases
NLL, RPS and bias compare matched cases. Group-restricted candidates are compared within their eligible population.
Check probability calibration
Compare nominal quantiles with observed coverage, then inspect coverage gaps and pinball loss.
Inspect the distribution behind a score
Compare forecast and observed PMFs/CDFs, or open a historical case with its uncertainty band.
