Indicator Lab Guide: How to Read Spearman Information Coefficient (IC), Quintile Monotonicity & DSP Formulas

Looking good on a chart and carrying measurable predictive information about future returns are two very different things. Every script you run in Indicator Lab is automatically audited across three statistical dimensions:

1. Spearman Rank Information Coefficient (IC)

Measures the rank correlation between the indicator at bar $t$ and forward returns at $t+k$ (+1, +5, +10, +20 bars):

• $|\text{IC}| < 0.02$ or $|t| < 2.0$: Indistinguishable from random noise.
• $|\text{IC}| \in [0.03, 0.06]$ with $|t| \ge 2.0$: Statistically significant, tradable edge.
• $|\text{IC}| > 0.08$: Strong predictive power across the tested regime.

2. Quintile Return Monotonicity (Q1 → Q5)

Historical bars are sorted by indicator value into five 20% buckets (Q1 lowest to Q5 highest). A genuine signal exhibits a clean, monotonic staircase from Q1 to Q5 rather than erratic mid-bucket spikes.

3. Example DSP Script: Ehlers Band-Pass Roofing Filter

// Removes both low-frequency spectral drift and high-frequency bar noise
rf = roofing_filter(close, 10, 48)
sig = ema(rf, 6)
plot rf = rf
plot sig = sig
plot cross_up = crossover(rf, sig) and rf < 0

Test this directly in Indicator Lab and share your IC decay tables below.