The honest test isn't "do high scores move more in absolute %" — the score deliberately picks low-volatility names, which always move less in raw terms. The thesis is expansion: does the move break out beyond the stock's own compressed band? That's Expand (realized ÷ expected) and Broke band (% exceeding ±1σ).
| Score bucket | bars | avg |move| | Expand | Broke band |
|---|---|---|---|---|
| ≥ 60 (coiled) | 7,032 | 6.5% | 1.19× | 47% |
| 30 – 60 | 14,632 | 6.3% | 0.96× | 37% |
| < 30 | 15,174 | 6.8% | 0.74× | 26% |
✅ coiled names expand beyond their own band more often — the squeeze thesis holds. Coiled bars broke their ±1σ band 47% of the time vs 26% for calm bars (Δ +21 pts). Raw absolute size barely differs (6.5% vs 6.8%; r = -0.04).
| State | bars | avg |move| | Expand | Broke band |
|---|---|---|---|---|
| squeeze ON | 2,611 | 6.7% | 1.10× | 45% |
| squeeze OFF | 34,227 | 6.5% | 0.90× | 34% |
Realized move landed inside the ±1σ band 65% of the time (theory ≈ 68%).
Overlapping forward windows make these autocorrelated — descriptive, not independent-sample stats. The score flags where a relative expansion is likelier, never its direction. Past behaviour does not guarantee future results.