This study examined the combined effects of teachers' instructional practices and students' reading-related affective engagement on predicting the high and low levels of elementary reading literacy from a linguistically and culturally comparative perspective. Data were based on 9748 students from 4 English-speaking and 3 Chinese-speaking education systems participating in the Progress in International Reading Literacy Study 2016. A mixed theory-based and data-driven approach was adopted. Four machine learning algorithms, specifically logistic regression, support vector machine, decision tree, and extreme gradient boosting, were simultaneously used to classify and predict high- and low-proficiency readers and to identify the most important factors for the ability separation. The findings showed that for both system groups, those factors together were sufficiently powerful to discriminate the readers and that the affective constructs, particularly students' self-concepts, played a predominant role. The Chinese-speaking systems, compared with their English counterparts, applied more effective pedagogies to nurture sophisticated readers.