## kernels.ConventionalReplay


Historical fits, origin-level outcomes, and explicit failures/exclusions.


Usage

``` python
kernels.ConventionalReplay(
    candidates,
    dates,
    fits,
    cells,
    errors,
    exclusions,
    loss_field,
    premium_field,
    units,
    segment
)
```


`fits` is keyed by (candidate name, information date). A failed candidate interval has no cell rows, and a reason in `errors` when recording errors. This prevents partial origin coverage from looking like a complete score.


## Parameter Attributes


`candidates: dict[str, ConventionalCandidate]`  

`dates: tuple[dt.date, …]`  

`fits: dict[tuple[str, dt.date], ConventionalFit]`  

`cells: pd.DataFrame`  

`errors: pd.DataFrame`  

`exclusions: pd.DataFrame`  

`loss_field: str`  

`premium_field: str`  

`units: str | None`  

`segment: dict`
