We analyze the winter of 2022-2023 in terms of forecast errors associated with atmospheric rivers (ARs) and precipitation in the Global and Integrated Forecast Systems (GFS and IFS) as well as two leading machine learning (ML) models (Graphcast and PanguWeather) over lead-times from one to ten days. This study utilizes the Method for Object-based Diagnostic Evaluation (MODE) from the Developmental Testbed Center on both integrated vapor transport (IVT) and precipitation. Using a simple distance metric, we relate precipitation objects to AR objects. This is done to ascertain the attributes of an AR system, which the correct forecast of, are most related to an accurate precipitation forecast. The attributes tested include the outer extent of an AR, the location of the core of an AR, the angle and curvature of an AR, the magnitude of the IVT within an AR, as well as the presence/proximity of an extratropical cyclone center. Two sub-periods are further analyzed to investigate the relationship between forecast skill and the large-scale flow regime. This work quantitatively documents both biases in AR forecasting and their relationship to precipitation forecast skill. Additionally, we document the relative skill of machine learning and numerical weather prediction models in this challenging forecast space.