PRISMOperator

Calibration

Decision feedback flywheel (CLAUDE.md §11 #1). Each operator-rated decision joins back to the opportunity that drove it; we compute Spearman ρ between each scoring factor and the operator's 1-5 success rating. With small n these numbers are exploratory — they tell you which way each factor probably points, not whether the relationship is statistically significant.

Decisions rated

0

Actionable

0

improve / launch / sample

Best predictor

Generated

4/30/2026

Per-factor Spearman ρ vs success rating

Trendn = 0ρ = n/ano signal
Saturation (1−)n = 0ρ = n/ano signal
Quality gapn = 0ρ = n/ano signal
Strategic fitn = 0ρ = n/ano signal
Total (composite)n = 0ρ = n/ano signal

Mean rating by opportunity confidence

Does our self-reported confidence track outcomes? If high-confidence decisions don't outperform low-confidence ones, the confidence model itself needs work.

low (≤0.3)

/ 5n = 0

medium (0.3-0.5)

/ 5n = 0

high (>0.5)

/ 5n = 0

No actionable rated decisions yet — buckets populate once outcomes are backfilled.

Interpretation

Spearman ρ in [-1, 1]. ρ > 0.3 = factor is a useful predictor. ρ near 0 = factor adds noise. ρ < -0.3 = factor is *inverted* (consider flipping its sign or weight). With n < ~10 these numbers are exploratory, not statistically significant.

Source: live · 2026-04-30.json · regenerated by python -m scoring.calibration.