I work in data analysis and this obsession over extreme spread is strange to me in this context.
ES is just an equation, it’s not a measure of accuracy. At best, it’s a proxy for consistency.
Case in point, you can have a single digit ES, and still miss your target entirely for a multitude of reasons.
What’s the proper statistical design for multiple, concurrent variables which may or may not have an association of causality?
Miltivariate Analysis?