How to spot outliers in wild, heavy-tailed data?

Extreme value analysis gets harder when data has heavy tails—a few massive outliers can break traditional ordering methods. Researchers introduce Polar Depth, a depth measure naturally suited to heavy-tailed multivariate data, expressed in polar coordinates that align with extreme value theory. Unlike halfspace depth, it better ranks the most extreme observations and converges provably to the theoretical limit. Tested on anomaly detection tasks.