Berg breadth and multi-index thrust indicator family — breadth thrust proxy
Rejected qm-tykd.3 · 2026-08-17
Do Berg breadth/multi-index thrust signals (advance/decline thrust, new highs/lows thrust, multi-index confirmation, breadth extremes) predict turning points beyond unconditional and event-appropriate baselines under causal lag/costs, and does prediction become a tradable overlay?
No. Primary HL Zweig breadth thrust proxy (10d HIGN/(HIGN+LOWN) low 0.40 -> high >=0.615 within 10d, 221 signals 2001-2026, 3.5% sparse) beats unconditional overall (mean fwd20 1.47% vs 0.92%, delta +0.55pp, hit 72.9% false 27.1% vs event high-only 0.73% vs 0.73% event n=4192) and vs event (+0.73pp) and retains at T+2 (-0.00pp), but fails frozen robustness: 2020-2026 delta -0.65pp hit 55.9% false 44% (regime failure vs 2001-2007 +2.07pp hit 84.4% and 2008-2019 +0.77pp hit 77.7%), adjacent thresholds/windows fail (high070 -0.10pp, window20 -0.06pp; no stable plateau), data alternatives unstable (S5FI variants fail early subperiod n=2-6; multi-index SMA fail -0.32pp; SPY up-day +0.09pp marginal), overlay 20d hold gross CAGR 1.61% vs SPY 9.22% delta -7.61pp (net 1.29% after 10bps, Sharpe 0.27 vs 0.56, maxDD -21% vs -55%, exposure 11.7%). Exact Berg thrust ratios/lookbacks withheld across all three reports — all results proxy-only; NYSE advance/decline daily history unavailable free tier, proxied via INDEX:HIGN/LOWN (6331 rows 2001+). 30k-indicator multiple-testing hazard dominates.
Provenance
Equity vs S&P (aligned, rebased to 1.0)
Aligned window, both rebased to 1.0 at start — Aligned window, both rebased to 1.0 at start for fair comparison · S&P dashed, experiment solid. X = Date, Y = Equity (rebased to 1.0).
Signal driver & thresholds (x-aligned)
Signal timeline (sparse triggers)
Triggers when overlay takes over base strategy
Stats
| CAGR | 0.00% | Sharpe | 0.00% |
|---|---|---|---|
| Sortino | 0.00% | maxDD | 0.00% |
Disposition
Rejected — breadth thrust proxy shows regime-dependent hit-rate that does not survive subperiod stability, adjacent-parameter robustness, or portfolio translation; no robust indicator to promote, no tradable strategy queued. All false thrusts and negative results retained.
Follow-on work
- No explicit follow-on; file new issue if needed
Errors & data gaps
NYSE advance/decline daily history unavailable free tier — proxied via INDEX:HIGN/LOWN highs/lows (6331 rows 2001+) which is index-level high/low count not true NYSE advance/decline; mismatch risk retained; true NYSE A/D thrust untestable as disclosed missing input Exact Berg breadth calculations, thrust ratios (e.g., advance/decline ratio), lookbacks and confirmation rules withheld across all three reports (zero verbatim numeric thresholds) — all thresholds/windows reported are inferred proxy neighborhoods kept distinct from verbatim disclosure DIA (DJI proxy) unavailable in data/market.sqlite (0 rows) — multi-index 4-index confirmation reduced to 3-index (SPY/QQQ/IWM); 30k-indicator multiple-testing hazard severe even for modest overall +0.55pp edge Sparse signal (3.5% frequency) — small subperiod ns (32 in 2001-2007, 59 in 2020+) widen confidence; regime non-stationarity dominates Equity curve is illustrative 20d holding overlay vs SPY buy-and-hold, not a recommended allocation — prediction vs portfolio kept separate per acceptance; overlay underperforms by 7.6pp
Follow-up
gh issue create --title "Follow-up Berg breadth and multi-index thrust indicator family breadt" --body "Follow-up to qm-tykd.3: Do Berg breadth/multi-index thrust signals (advance/decline thrust, new highs/lo" --label "priority:p2" --label "pipeline:research" --label "agent:ready"