Aqyl Battle: item-level difficulty statistics of a Kazakh-Russian quiz learning game (no texts, no learner data)

Open data from Aqyl Battle, a Kazakh-Russian mobile quiz-based learning game, accompanying the article 'Text-Based Item Difficulty Estimation, IRT Calibration and Learner Performance Prediction on a Kazakh–Russian Gamified Learning Platform' (IJACSA). Item-level statistics for 30,444 calibrated quiz items (category, language, translation links, attempt counts, logit difficulty, leakage-safe splits), test-set predictions of 36 difficulty models, IRT parameters, duplicate flags and cold-start predictions for 416 tournament items, and category-level Elo difficulty. Item texts and learner-level records are not included: they belong to the operator and include data of minors, and are available on request under a data-use agreement. Platform identifiers are replaced by random ids. A verification script recomputes the key results of the article from these files.