Evaluation

class omnicast.Backtester(horizon=1, initial=None, step=1, metric='rmse')

Bases: object

Expanding-window rolling-origin backtest.

Parameters:

• horizon (int)

• initial (int | None)

• step (int)

• metric (str)

run(model, y)

Evaluate model on expanding-window temporal splits of y.

Returns:

Fold scores (cutoff, score, n_train) in .scores, per-step out-of-sample predictions in .predictions, plus .summary()/.plot().

Return type:

BacktestResult

Raises:

ValueError – If initial and horizon leave no validation observations.

Parameters:

• model (BaseForecaster)

• y (Iterable[float] | Series)

omnicast.backtest(model, y, horizon=1, initial=None, step=1, metric='rmse')

Evaluate a forecaster on expanding-window temporal splits.

One-shot wrapper over Backtester(...).run(model, y).

Return type:

BacktestResult

Parameters:

• model (BaseForecaster)

• y (Iterable[float] | Series)

• horizon (int)

• initial (int | None)

• step (int)

• metric (str)