Numerical discrimination is a primary measure of the acuity of children’s approximate number system (ANS). ANS acuity is associated with key developmental outcomes such as symbolic number skill, standardized test scores and even employment outcomes. The current study examines the factors that contribute to children’s performance on non-symbolic numerical discrimination tasks. The current study evaluates the contribution of absolute value in children’s numerical discrimination, and how that contribution may change during development. We use a combination of behavioral and computational results to illustrate a U-shaped developmental change in the factors that predict numerical discriminability. Computational modeling based on the neural coding of numerical perception demonstrates why the reported behavioral data is expected. The novel inclusion of absolute value as a predictive factor in children’s numerical discrimination suggests reevaluation of connections between numerical acuity and educational outcomes.