The role of sports games in the research and application of children's intellectual development is very important, but there is a problem of inaccurate evaluation of results. Multi-marker learning cannot solve the research and application problem of children's intellectual development in sports games, and the research application is unreasonable. Therefore, this paper proposes a multi-example multi-marker learning for research and application analysis of children's intellectual development. First of all, mathematical theory is used to research and apply sports games, and indicators are divided and reduced according to the research and application requirements of children's intellectual development Interference factors in the research application of early childhood intelligence development. Then, the research application of educational theory to children's intellectual development in sports games forms a research and application program for children's intellectual development, and is right the research and application results of children's intellectual development are comprehensively analyzed. MIML simulations show that multi-example multi-label learning is applied to sports games under certain criteria The research application accuracy of children's intelligence development and the research application time of children's intelligence development are better than multi-marker learning.