The sum score on a psychological test remains a central tool in psychometric practice, despite some psychometricians' belief that it should be replaced by latent variables. This paper argues that the sum score stochastically orders the latent variable in many item response models and is mathematically justified in ordinal use. Classical test theory (CTT) provides lower bounds for reliability, which are close to true reliability under reasonable conditions. The sum score's value lies in its ability to predict practical events and behaviors. While modern models like IRT have their merits, CTT offers impressive contributions with minimal assumptions. The sum score is not obsolete; it has desirable properties and can be superior to latent-variable models in some cases. The paper discusses the sum score's relationship with the latent variable, its use in network models, and its reliability estimation. It concludes that the sum score remains a useful tool in psychometrics, despite criticisms. The paper also highlights the importance of understanding CTT and its role in reliability estimation, and the need for a balanced view of both CTT and IRT. The sum score is a robust ordinal estimator of the latent variable and can be used in complex network models. The paper emphasizes the need for a reevaluation of the sum score's role in psychometrics.The sum score on a psychological test remains a central tool in psychometric practice, despite some psychometricians' belief that it should be replaced by latent variables. This paper argues that the sum score stochastically orders the latent variable in many item response models and is mathematically justified in ordinal use. Classical test theory (CTT) provides lower bounds for reliability, which are close to true reliability under reasonable conditions. The sum score's value lies in its ability to predict practical events and behaviors. While modern models like IRT have their merits, CTT offers impressive contributions with minimal assumptions. The sum score is not obsolete; it has desirable properties and can be superior to latent-variable models in some cases. The paper discusses the sum score's relationship with the latent variable, its use in network models, and its reliability estimation. It concludes that the sum score remains a useful tool in psychometrics, despite criticisms. The paper also highlights the importance of understanding CTT and its role in reliability estimation, and the need for a balanced view of both CTT and IRT. The sum score is a robust ordinal estimator of the latent variable and can be used in complex network models. The paper emphasizes the need for a reevaluation of the sum score's role in psychometrics.