What is weighted relative error?

Weighted relative error gives more importance to application tiers or samples that matter more for allocation decisions.

Plain meaning

Relative error measures proportional deviation between prediction and actual result. Weighting makes key tiers contribute more to the overall metric.

Why it matters

Not every tier matters equally. One-lot, A-tail and B-head tiers often matter more for real subscription decisions.

How to read it

Lower is better, but read it together with P90, sample size and tier-level deviations.

How it appears here

HK IPO AI uses weighted relative error on the allotment page to summarize historical allocation prediction error without treating every tier as equally important.

Common mistakes

Thinking weighted error is a simple average of all tiers.
Looking only at the headline error without key-tier errors.
Ignoring how very hot or very cold IPOs can change the error distribution.