There are only two "best practices," and they tug in opposite directions: 1. compression (number of bits needed to encode) 2. simplicity (ease of understanding, by the mind) Perhaps the best example in all of science is Geometric Algebra's rewriting of Maxwell's four equations into the single expression: $\nabla F = J$ where $\nabla$ is a simple but unusual kind of derivative that encodes both the divergence and curl using something called the Clifford Product.