This paper proposes a continuation-based framework for persistence in constrained sequential systems. Rather than depending primarily on current state quality, the framework proposes that persistence depends more strongly on whether coherent future continuation still remains behaviorally accessible. The framework is tested using large-scale chess data and engine-derived continuation-space proxies. Across multiple analyses, continuation-related variables predicted persistence more strongly than raw position evaluation alone. The results suggest that persistence may depend less on instantaneous state quality than on the structure of remaining future accessibility.
Zarko Cosovic (Tue,) studied this question.