Physical entanglements between polymer chains demonstrably enhance the modulus, strength, and toughness of elastomers and gels. However, it remains difficult to directly relate entanglement micromechanics to their macroscopic mechanical benefits. Experimentally investigating entanglements is difficult due to their nanoscale sizes, subsurface locations, and chemical indistinguishability from their surroundings. Computationally mapping structure-property relations remains difficult due to the high cost of predicting macroscale properties using physics-based models that enable direct entanglement observation, such as coarse-grained molecular dynamics (CGMD). Furthermore, entanglements are transient, configuration-dependent features without clear quantitative definitions. To address this ambiguity, we introduce an approach that quantitatively defines local entanglements along the backbones of simulated polymers based on the Gaussian Linking Number. We further introduce a geometric center of entanglement and verify that it represents the position through which entropic chain forces are transmitted using Kremer-Grest CGMD simulations. Unlike existing approaches, which output a single linking number for pairs of chains, our approach identifies the multitude of load-transmitting, inter- and intra-chain entanglements along a polymer’s backbone. Towards bridging scales, we introduce a topological distillation algorithm that converts entangled CGMD networks into representative discrete network models (DNMs), representing entanglements as vertices and primitive paths between them as load-transmitting edges. Our algorithm generates DNMs that reproduce small-strain virial stress predictions of the Kremer-Grest model with upwards of 97% reduction in computational cost, verifying the physical accuracy and cost benefits of this distillation approach. Moving forward, this distillation procedure will facilitate physics-based, predictive modeling of entangled network mechanics, from polymers to architected metamaterials.