Competitive Judo can be interpreted as a natural laboratory of adaptive behaviour, where decision-making, structural constraints and dynamic perturbations continuously interact within a complex system.

This work proposes a conceptual and methodological framework that frames Judo performance analysis within a perspective grounded in Engineering Resilience, with particular reference to socio-technical and economic–organizational systems.

Judo is modeled as a complex adaptive system, characterized by interacting subsystems, state transitions, feedback loops and dynamic reconfiguration under perturbations. Fundamental technical–tactical phases are interpreted as observable system states, while competitive interactions are treated as structured perturbations capable of triggering adaptive responses.

Within this perspective, the classical Judo sequence 'Kuzushi–Tsukuri–Kake' may be interpreted as an operational analogue of adaptive control logic: detecting imbalance, configuring an appropriate response and executing action at the right moment.

The proposed modeling architecture integrates 'Design Structure Matrix' (DSM), 'Full Random Effects Model' (FREM) and Machine Learning methodologies within a unified analytical framework.

• DSM represents and dynamically updates interdependencies among functional components (processes, decision nodes, sensing modules and agents), enabling structural reconfiguration analysis.
• FREM captures hierarchical and cross-context variability, formalizing inter-athlete and situational heterogeneity within a statistically coherent framework. Supervised and Reinforcement Learning techniques are introduced to identify recurrent configurations of effectiveness, infer adaptive decision policies and iteratively adjust system parameters and interdependency weights.

The computational pipeline (processing chain) includes:

1. multimodal data acquisition (video analysis and inertial sensing),
2. extraction of kinematic and dynamic features,
3. structural modeling through DSM,
4. hierarchical statistical estimation via FREM, and
5. learning of decision policies through supervised and reinforcement approaches.

The architecture is then conceived as a transferable modelling framework for adaptive Risk Management and Decision Support in complex environments characterized by uncertainty and dynamic perturbations. The resulting system enables formal representation of adaptation mechanisms and feedback-driven reconfiguration processes, mapped onto core Engineering Resilience constructs — shock absorption, functional redundancy, resource reallocation, learning and adaptive governance.

The contribution is primarily conceptual and methodological, rather than empirically validated on large-scale datasets. Its objective is to provide a reproducible and critique-ready modeling architecture capable of fostering interdisciplinary integration across Motor Sciences, Systems Engineering, Data Science and Organizational Economics, contributing to the development of a shared analytical language for resilience analysis grounded in observable system dynamics.