Скачать или смотреть Quantifying the Win Rate and Gold-Equivalent Value of Atakhan Using Causal Machine Learning

We evaluate the causal impact of securing Atakhan, Bringer of Ruin—a new epic neutral objective in League of Legends—on team win probability. Using over 77,000 event-level observations from North American Challenger-tier matches (patches 15.14–15.15), we apply a Causal Machine Learning approach combining Doubly Robust (DR) estimation with Conditional Average Treatment Effect (CATE) analysis. Controlling for gold difference, turret count, and event timing, we find that securing Atakhan increases win probability by an average of 45.79 percentage points, peaking at an increase of 52.24 percentage points in the 22–26 minute window. A logistic regression–based gold-equivalent conversion estimates its median strategic value at approximately 11,701 gold—exceeding the combined worth of multiple major turrets. These results position Atakhan as a decisive late-game win condition and illustrate the utility of causal inference methods for measuring the true in-game impact of high-value objectives in competitive eSports.