Normalized Discounted Cumulative Gain (nDCG)

A position-sensitive ranked-retrieval metric that sums graded relevance gains across result positions with a logarithmic discount by rank (commonly using 1 / log2(rank+1)), producing Discounted Cumulative Gain (DCG), and then normalizes DCG by the ideal DCG (IDCG) for that query to yield nDCG in the range [0,1]. It supports multi-level (graded) relevance judgments.