Z-score calculator

Reading the two derived figures

The percentile assumes a normal distribution. That assumption is convenient and, for the quantities Moonboard standardises, not accurate. Interaction counts and turnover are heavy-tailed, so the real share of assets below a given z-score differs — often substantially — from what the normal curve implies. Treat the figure as an orientation, not a measurement.

The COR mapping shows what COR would produce if this were the mean z-score across its five inputs. Notice how compressed it is: a z-score of +1 — a genuinely strong result — maps to 66.7, and reaching 100 would require +3 on the average of five standardised metrics, which effectively never happens.

Why standardise at all

Standardisation lets quantities in different units be combined. Sentiment runs 0–100, interactions run into the millions, turnover is a small fraction. Summing them directly means the largest one dominates — precisely what happens in SMI, where interaction counts carry over 99 % of the composite despite a nominal weight of 0.4.

Converting each to a z-score first puts them on a shared scale where a weight means what it says. That is the design difference between COR and the other multi-factor metrics.