Brazilian Rankeds Game League of Legends

This dataset contains 11,007 ranked Solo/Duo League of Legends matches collected from the Brazilian server (BR1) through the Riot Games API. For each match, the dataset includes per-minute timeline information transformed into team-level aggregate features, such as gold, gold difference, kills, experience and level proxies, creep score, vision statistics, elite objectives, towers destroyed, and first-objective indicators. The dataset was designed to support research on early win prediction in League of Legends, especially comparisons between snapshot-based, flattened temporal, and sequential representations at 10, 15, 20, and 25-minute decision cutoffs. The target variable is the binary match outcome, indicating which team won. Duplicate matches and prematurely ended games, such as remakes or early surrenders, were removed. Only information available up to each decision cutoff is used, avoiding label leakage from later match events or end-of-game records. No personally identifiable player information is retained; the dataset contains only aggregate gameplay statistics and anonymized match outcomes. This release supports replication of experiments comparing tabular and sequential machine learning models for cost-aware early outcome prediction in MOBA games.