Every active scenario carries a probability between 0 and 100%. Here is the calculation, in order.
Step 1 — Signal position
For each signal in the map, Numen computes where the current value sits in the trailing 12-month percentile. A signal at its 12-month high lands at 1.0; at its low, 0.0; at the median, 0.5.
Step 2 — Alignment
Position is flipped by direction. If a signal is marked Positive (rising favors the scenario), alignment equals position. If marked Negative (falling favors), alignment equals (1 − position).
Step 3 — Strength coefficient
Each signal's strength (0.0–1.0) is multiplied by 16 to produce a coefficient. A strength of 1.0 carries 16× a strength of 0.1. The exponential-feeling weight comes from the multiplier, not from any nonlinearity in your input.
Step 4 — Staleness damping
If a signal has not been observed for more than 14 days, its alignment dampens toward 0.5 (neutral) on a linear schedule. After 30 days of no data, the signal contributes only neutral weight.
Step 5 — Weighted score
Sum (coefficient × alignment) across all signals, divide by total coefficient, get a raw score in [0, 1].
Step 6 — Logistic calibration
Push the raw score through a logistic function with k=4 to get the final probability in [0, 100]. The logistic squashes extremes so very-near-0 and very-near-1 raw scores are tempered.
Step 7 — Confidence
Confidence is high if you have 6+ fresh signals with no single signal carrying more than 40% of total weight. Medium if 3+ signals, mostly fresh, no signal above 55%. Otherwise low.