@ Drift Hunters zz0.ku4iybxkhgg zz Yes — what you’re describing is not a new problem. It’s basically a combination of well-known techniques:
  1. Percentile / quantile estimation
    Your “L where P% of points are below L+E” is essentially a high percentile (e.g. 95th, 99th percentile).
    Instead of inventing L, you can compute:

L = quantile(data, P)

This is standard and implemented everywhere (NumPy, SciPy, pandas, Apache...