Grounded in flow theory, numerous researchers and practitioners have proposed that the relationship between game difficulty and engagement follows an inverted U curve, with optimal engagement at a ‘goldilocks’ midpoint where game difficulty ‘matches’ player skill. Existing empirical studies have found mixed support for this relationship, but also show high variance in operationalising the proposed relationship: operationalising game difficulty as e.g. winning odds, average scores, winning rates, level manipulations, use or non-use of dynamic difficulty adjustment, etc. operationalising game engagement as perceived enjoyment, perceived need satisfaction, play time, etc. Maybe even more importantly, contrary to flow theory, game difficulty is typically theorised and operationalised as a fixed, subject-independent construct, rather than as a subject-environment ratio or relationship. That is, optimum ‘medium difficulty’ is construed as a population-level aggregate medium, rather than as an individual-level match with skill. One likely reason prior research has found mixed results is that it did not operationalise difficulty on an individual level, producing higher variance in the data overall and biased results due to unrepresentative samples with non-normally distributed playing skills for the game under study.