Berg price and turning-boundary indicator family — price boundary proxy
Rejected qm-tykd.1 · 2026-08-17
Do Berg price/turning-boundary signals (price extremes, boundary breaks, reversal levels, support/resistance envelopes as sparse turning-point signals) predict turning points beyond unconditional and near-boundary event baselines under causal lag/costs, and does prediction become a tradable overlay?
No. Primary SPY rolling 252d percentile <=5th proxy (389 signals 1994-2026, 4.77% sparse) beats unconditional overall (mean fwd20 T+1 1.45% vs 0.92% delta +0.52pp hit 62.7% false 37.3% vs near-boundary event pct5-10th -0.51% hit 49.3% delta +1.96pp) and retains at T+2 (-0.04pp to 1.41%), but fails frozen robustness: 2008-2019 delta -0.70pp hit 61.0% false 39% (regime failure vs 1993-2007 +0.18pp hit 56.6% and 2020+ +4.42pp hit 79.7%), adjacent thresholds/windows fail (pct10th -0.19pp, env -5% -0.10pp; no stable plateau; pct1st +2.47pp and env -10% +1.13pp sporadically beat primary), window 63/126 beat 252 (+0.87-0.98pp), overlay 20d hold gross CAGR 0.49% vs SPY 10.80% delta -10.31pp (net 0.49% after 10bps Sharpe 0.10 vs 0.64 exposure 10.6% 38 transitions) and env -8% 1.97% vs 10.80% delta -8.82pp. Exact Berg boundary thresholds/lookbacks withheld across all three reports — all results proxy-only; SPY OHLCV close_only 1993+ 8428 rows. 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 — price/boundary proxy shows regime-dependent hit-rate that does not survive subperiod stability, adjacent-threshold/window robustness, or portfolio translation; no robust indicator to promote, no tradable strategy queued. All false boundary touches and negative results retained.
Follow-on work
- No explicit follow-on; file new issue if needed
Errors & data gaps
Exact Berg price-boundary formula, thresholds (e.g., envelope width, percentile cutoff, support distance), lookbacks (63/126/252), reversal confirmation rules and asset scope withheld across all three reports (zero verbatim numeric thresholds) — all thresholds/windows reported are inferred OHLCV-derived proxy neighborhoods (rolling percentile, SMA envelope, support distance) kept distinct from verbatim disclosure; native Berg 30k series proprietary/unavailable SPY OHLCV via data/market.sqlite close_only (8428 rows 1993-01-29..2026-07-24, effective 1994-01-28 after 252d warmup, 8156 valid days) — Volume=0 close_only sufficient for price family; QQQ/IWM similar; no NYSE volume/breadth license needed 30k-indicator multiple-testing hazard severe even for modest overall +0.52pp edge with ~12 implicit threshold/window trials; hit-rate 62.7% below unconditional 65.3% (more losers despite larger winners); 2008-2019 delta -0.70pp regime failure dominates Equity curve is illustrative 20d holding overlay (pct5 within 20d → long SPY else cash, T+1 lag, fee_bps=10 negligible at 0.6/yr) vs SPY buy-and-hold, not a recommended allocation — prediction vs portfolio kept separate per acceptance; overlay underperforms by 10.3pp
Follow-up
gh issue create --title "Follow-up Berg price and turning-boundary indicator family price boun" --body "Follow-up to qm-tykd.1: Do Berg price/turning-boundary signals (price extremes, boundary breaks, reversa" --label "priority:p2" --label "pipeline:research" --label "agent:ready"