An accurate model for the lift force on a baseball is important for several applications. The precision of previous models has been limited by the use of small samples of measurements acquired in controlled experiments. The increased prevalence of ball-tracking radar systems provides an abundant source of data for modeling, but the effective use of these data requires overcoming several challenges. We develop a new model that uses this radar data and is constrained by the physical principles and measurements derived from the controlled experiments. The modeling process accounts for the uncertainty in different data sources while exploiting the size and diversity of the radar measurements to mitigate the effects of systematic biases, outliers, and the lack of geometric information that is typically available in controlled experiments. Fine-grained weather data is associated with each radar measurement to enable compensation for the local air density. We show that the new model is accurate enough to capture changes in lift due to small changes in surface roughness which could not be discerned by previous models.