Insights and guides to help you master League drafting
An internal study on 10,000 ranked games measuring suggestion lift at every pick position, under two baselines: random pick order vs a player who already saves their strongest champion for counterpick.
Head-to-head against LoLTheory on ~33,000 held-out ranked matches, with a qualitative example showing where linear pairwise models miss team-wide damage-profile effects.
Comparing LoLDraftAI against iTero's drafting tool. A qualitative walkthrough of where pairwise-statistics models miss the compositional story, a feature coverage table, and a 100-draft head-to-head tournament.
A head-to-head comparison against DraftGap on ~33,000 ranked matches played after both tools were frozen, with a qualitative example showing where statistical draft analysis falls short.
A detailed explanation of how LoLDraftAI works.
A correction regarding the previously claimed 62% accuracy statistics that were affected by overfitting issues.