Purpose: The purpose of this paper is to propose and demonstrate an evidence-based scenario design framework for assessment-based computer games. Design/methodology/approach: The evidence-based scenario design framework is presented and demonstrated by using BELLA, a new assessment-based gaming environment aimed at supporting student learning of vocabulary and math. BELLA integrates assessment and learning into an interactive gaming system that includes written conversations, math activities, oral and written feedback in both English and Spanish, and a visible psychometric model that is used to adaptively select activities as well as feedback levels. This paper also reports on a usability study carried out in a public middle school in New York City. Findings: The evidence-based, scenario design framework proves to be instrumental in helping combine game and assessment requirements. BELLA demonstrates how advances in artificial intelligence in education, cognitive science, educational measurement, and video games can be harnessed and integrated into valid instructional tools for the classroom. Research limitations/implications: This paper provides initial evidence of the potential of these kinds of assessment-based gaming tools to enhance teaching and learning. Future work involves exploring student learning effects in randomized controlled studies and comparing the internal assessment models to more traditional assessment instruments. Originality/value: BELLA is the first step toward achieving engaging, assessment-based, gaming environments for a variety of Science, Technology, Engineering and Math (STEM)-related areas with explicit support for English language learners. (Contains 8 figures.)