running the model on past elections where you already know the outcome — is the primary tool for this validation.