Indicators

risk_off iff TIP avg(1/3/6/12) momentum <= 0

risk_off unless both TIP and SPY avg(1/3/6/12) momentum > 0

risk_off unless all canary 13612W scores > 0

crash_fraction = min(1, count(canary_13612W <= 0) / 2); active iff crash_fraction > 0

• Continuous CF; booleanization uses any CF>0 (partial risk-off counts as active).

CBF from count(offensive 13612W <= 0) / breadth_b; active iff CBF > 0

• Aggressive and Balanced share the rule family but differ in universe and B; gauge producer must not double-count without an explicit choice.
• inputs.tickers left empty because universes differ by variant.

If UNRATE YoY at prev month <= 0 → risk_on; else risk_off iff any canary 13612W <= 0

risk_off iff ue_recession_signaled AND SPY < 10-month SMA

risk_off iff recession_signaled AND SPY < 10-month SMA

risk_off when CLI diffusion / asof signal fails DI threshold (classic inclusive DI>=0.5 risk-on; Enhanced strict >0.5)

• Classic mid-month vs Enhanced monthly/mid-month share the CLI gate but differ on threshold inclusivity and post-gate asset selection.
• Producer must pick asof vs diffusion consistently with each strategy variant.

risk_off iff SPY 12m return <= BIL 12m return

cp_fraction from breadth of positive MOM; active iff cp_fraction > 0

• PAA vs PAA-CPR differ only in CP sleeve selection (IEF vs max(IEF,BIL)), not the CP fraction itself.

cp_fraction from count(risk zi > 0); active iff cp_fraction > 0

invest_fraction from canary_invest_fraction; risk_off_fraction = 1 - invest_fraction; active iff invest_fraction < 1

• Booleanization uses invest_fraction < 1 (any canary failure). Producer should also expose the continuous risk_off_fraction.

risk_off iff n_bad_13612w >= 1 OR n_bad_1m >= 3

• Inferred from AS allocations/returns; not an author-disclosed formula.

risk_off_pct = 1 - risk_on_pct; active iff risk_off_pct > 0

• inputs.tickers omitted (full Vitral universe is large); producer should pull VITRAL_UNIVERSE from module.

risk_off iff ue_recession_signaled(..., require_rising=True / Bond UI1 path)

If no COMP trigger → risk_on; if any trigger → risk_off iff SPY < 200-day SMA

• Reverse-engineered COMP thresholds; proprietary/black-box caveats remain on AS SPY-COMP pages historically.

risk_off iff max(SPY_score, SCZ_score) <= 0 (neither equity clears absolute momentum)

risk_off iff any of the five 3m returns is negative

risk_off iff SPY < 10-month SMA

risk_off when 10y <= 3m (original); amended treats YC as risk_on after 12 consecutive inverted months

• Amended variant can flip the same YC observation to risk-on after the 12-month cap; producer should emit both or document which is gauged.

risk_off unless TIP>12m SMA AND div_yield>=1.6% AND no YC inversion in lag window 7..15

risk_off when SPF recession averages worsen (spf_recession_risk_on is False)

risk_off (macro sleeve defensive) iff mcc_macro_recession_signaled (>=2 of 3)

• Only one of four MCC sleeves; not a whole-portfolio binary. Still useful as a distinct macro indicator.

cp_fraction = count(ri<=0)/N; active iff cp_fraction > 0

• Offensive vs defensive scoring differs (zi formula); CP count still uses ri sign.

risk_off iff zero risk assets are within near_pct of 12m high

• Partial sleeve drop-outs are not boolean risk-off; only the all-off state is inventoried here.

WSHOSHO weekly→daily forward-filled ROC5 >4.8% with publication lag L=7 and 20-day fixed holding window. Selected as best validated version from qm-4bx.19 grid; alternative filters (zscore, level-deviation, acceleration, YoY/WoW) evaluated head-to-head and retained ROC as most robust (negative result).

USI:UVOL / USI:DVOL daily ratio >22 with 20-day fixed holding window. Selected as best validated version from qm-4bx.17 full grid (aggregation daily vs sum5 5-day vs MA {SMA,EMA,WMA,Wilder,HMA}×{3,5,10,20}, thresholds {8,12,15,22,30,40}+expanding 90/95th quantile/σ, hysteresis 1-3, holding {5,10,20,40,60} fixed/decay/retrigger) evaluated marginally vs 3 bases (SPY B&H, SMA200 risk-off, 60/40 balanced) with 10bps+2×, subperiods, 60m walk-forward, +1 bar lag, synthetic-vs-native, falsification, brittle-22 flag; sum5 at 22 never fires (max 5.08), MA smoothing at 22 collapses to 0, threshold brittle flagged — daily raw 22 retained as scripted anchor, hw20 balances capture vs stale (40/60d best 17.5/18.6 vs 13.39 but monotonic overfit risk).

INDEX:S5FI % S&P 500 above 50dma, HMA20 crossing up through 8.5% after washout or crossunder 0% (mega washout), absolute thresholds, crossover trigger, 20-day fixed holding. Selected as scripted anchor from qm-4bx.18 full grid (not cherry-pick; low10_mega0 9.55 >8.01 but anchor balances brittleness and walk-forward robustness).

Philadelphia Fed SPF Anxious Index (mean probability of GDP decline next quarter, RECESS2) with disclosed threshold 28.2% as recession-warning level. Quarterly native since 1968Q4, never-revised, published via SPF release months 2/5/8/11 (~6-week lag) with monthly forward-fill to daily calendar; active when anxious >=28.2 (risk warning, not buy impulse). Selected as disclosed anchor from https://recessionalert.com/recession-forecasting-with-the-anxious-index/ page sha256:91b3792b9d00ecf469c3331c00ed642ab51bb5588deacb3eaec868ffb1cba053; validation reproduces AUC 0.94 (0.944 pre-2012) vs NBER months.

RecessionAlert U.S. Monthly Leading Economic Index (USMLEI V2, -100 to +100 continuous gauge scaled from 21-23 standardized leading components, 6-9 month outlook). Month-end observation with ~1-month publication lag (Reports Centre PDF, e.g., Oct 2022 published Nov 30). No disclosed discrete buy/sell threshold; negative print (USMLEI < 0) is descriptive caution band only, not a trading rule. Published from https://recessionalert.com/usmlei-oct-2022-published/ page sha256:dbea5c54e0bcff96b63e494f0c063674851101fd9b5d371fed2f0f078ef2aaf4 and MonthlyData.xlsx DATA col 16 USMLEI V2 sha256:435d142f7e8851a740c78a3d24db6bb15930b21a861cb5520d0a494e6575dced (698 months 1968-05→2026-06; Oct 2022 -18.4 vs page -19.7). Continuous overlay with caution band while <0.

Buy-only Class A/B Great Trough impulses replayed exactly from the SHA-verified calculated WeeklyData.xlsx snapshot (83 unique signal sessions, 2015-03-26 through 2026-08-14). The 2014 disclosed rule arms Class B below 27.4 for 5 trading days or Class A below 10 for 10 trading days, then buys on a punch above 80 after at least 4 days below 80. The proprietary breadth smoothing constant and exact pre-2015 history are unavailable, so this publication does not reconstruct them. At +1-bar execution, the combined signal trails the fixture-span unconditional S&P baseline at every disclosed 37/83/119/150-session horizon; publication is for reproducibility and inspection, not an efficacy or strategy endorsement.

• Raw-formula reconstruction is fail-closed: the breadth smoothing constant is proprietary.
• The workbook supplies exact calculated GTR fields only from 2015-03-26; earlier claimed history is chart-only and is not inferred.
• Workbook marker codes 1 versus 2 are undocumented; only positive versus zero is normalized.

Buy-only Zweig Breadth Thrust Redux (ZBT-A) impulses replayed exactly from the SHA-verified calculated WeeklyData.xlsx snapshot (12 ZBT-A markers, 2004-05-21 through 2022-10-04; 15 classic ZBT and 89 rapid ZBT-B markers retained for audit). The page- disclosed classic ZBT rule is 10-day EMA of NYSE Advances/(Advances+Declines) from <0.40 to >0.615 within 10 trading days; the Redux ZBT-A rule is described only as stricter lower + relaxed upper + extended window (~4x frequency, 13-chart signals) without numeric disclosure and is therefore workbook-replayed not reconstructed. At +1-bar execution all 37/83/119/150-session horizons beat the fixture-span unconditional S&P baseline (37d mean 0.018 vs 0.012 win 66.7% vs 66.5%, 83d 0.045 vs 0.027 win 83.3% vs 70.3%, 150d 0.102 vs 0.049 win 83.3% vs 74.7%); extra lag T+2 is stable. Publication is for reproducibility and inspection, not a strategy endorsement.

• Exact numeric ZBT-A lower/upper thresholds and window are undisclosed beyond qualitative stricter/relaxed/extended; replayed from workbook ZBT-A BUY.
• Exact numeric ZBT-B upper threshold is undisclosed (described only as relaxed vs 0.615 with 12-day window).
• Proprietary 10-day EMA construction and NYSE Advances/Declines source vintage are undisclosed.
• Pre-2004 and pre-1945 history is chart-only (images) without machine-readable series and is not inferred.
• Workbook snapshot revision/vintage policy is unknown; current snapshot only.

AAII bearish % beyond 95th percentile of trailing 2 years — retail fear extreme (contrarian BUY context). Threshold reconstructed (values not published by Berg).

AAII bull-bear spread beyond 5th percentile of trailing 2 years.

5-day advance/decline ratio between 0.35 and 0.39 near a low — 'reverse breadth thrust' from the first 1957 signal. NYSE A/D when available, else up/down return-ratio proxy.

S&P makes 252d high while cumulative A/D line stays >=2% below its own high — narrowing advance (Mar 1987 template).

• Computed series unavailable; cited historical dates are retained.

The combined 2-component model — 6 occurrences (1982, 1997, 2009x3, 2011), one-year gains +19.7% to +49.9%, low always held. Needs constituent breadth for the 95% leg; NASDAQ ROC leg lives in signals_roc.ixic_roc10_pos10.

• Computed series unavailable; cited historical dates are retained.

CFTC noncommercial S&P futures net position extremes as contrarian gauge. Free source identified (cftc.gov fut_fin_txt) but not yet ingested.

• Required series not freely available; registered with missing_data.

Two consecutive NASDAQ gains >=1.90% with three consecutive S&P gains >=1.60% — 5 precedents before Apr 24 2025; low never re-breached.

S&P -12%+ decline, low held 2 days, then +3% off the low on NYSE higher volume — 4 occurrences (1962, 2003, 2010, 2022), +21% to +39% next year.

Decline >=18% reaching a fresh 60-day low (thresholds list tier).

Deep-decline tier capturing the 1962 template (-27.97% into a 60-day new low).

GFC-tier decline (Berg cites -45% into the 2009 low).

Dow Jones Utility Average peaks on record volume while S&P near highs — the Jan 22 1987 divergence that preceded the crash.

'Wait till the market declines 4%... and get out then' — earliest exit tier.

7% exit variant — backtested by Berg; fires more than once a year.

THE MB Edge sell rule: exit to T-bills on an 8% closing drawdown; ~1 round trip per 1.25 years. Re-entry on first buy signal.

Equal-weight S&P 6-day ROC / 18-day ROC >= 105% — 17 occurrences, low never broken, median +32.5% max 12-mo gain. Uses SPXEWI/RSP when available, else cap-weight S&P proxy.

Sentiment turning MORE bullish into a tightening cycle — complacency hallmark before the 1987 crash.

Gap up held all day, gap down the next day — first in RUT history on Jan 22-23 2026; laggards-leading top marker.

NASDAQ 10-day ROC >= +10% — thrust component of the 6-occurrence breadth thrust model (1982, 1997, 2009x3, 2011); low always held.

NASDAQ Composite 3-day rate of change <= -7% — panic acceleration; one of the three simultaneous Apr 4 2025 extremes.

Money-supply growth slowing vs industrial production + commodity prices — preceded ~80% of 10%+ drops since 1960 (Berg, Apr 2021 warning). FRED M2SL/INDPRO/PPIACO yoy; FRED currently unreachable from this host, so the computable leg is pending a re-fetch.

• Computed series unavailable; cited historical dates are retained.

'Act on the first buy signal' — the 55 BUY legs isolated for bottom-call comparison against reconstructed signals.

The published binary model history: BUY = 100% S&P 500, SELL = 100% T-bills. 110 legs, 1957-10-21 to 2026-03-27, extracted verbatim from the site's RSC payload. Ground truth for alignment scoring.

Calendar volatility windows from Paul Macrae Montgomery's cycle work (Berg's 'greatest single influence'). Cited windows: 1987-08-24→11-05, 2024-08-15, 2026-01-30. Historical events only — cycle tables proprietary.

NAAIM member equity exposure at extreme (paywalled since 2026-08); placeholder with citation.

• Required series not freely available; registered with missing_data.

NASDAQ advance-decline line down 8 of 10 days with weak 5-day — Feb 5 2026 panic cluster component (1997 analog +39%).

NASDAQ-100 5-day volume extreme not seen in >=200 days — Apr 4 2025 triple. Needs NDX volume (Yahoo ^NDX carries it; TV index volumes are zero) — currently data-gapped.

• Computed series unavailable; cited historical dates are retained.

Net up volume as % of total NYSE volume, 5-day average <= -45% — the 1957-10-21 first-signal condition.

5-day net upside volume beyond the -30%/+35% noise zone (5.74%/5.26% tails of 22,169 obs) — rarity-principle reconstruction.

Equity-only put/call beyond 95th percentile of trailing 2 years.

Total put/call ratio beyond 95th percentile of trailing 2 years — hedging panic; contrarian BUY context (CBOE free CSV 2003+).

Dynamic-hedging/portfolio-insurance dominance turned ordinary declines into the Oct 1987 cascade. Historical event only.

TRIN zones acting as 'reflecting boundaries' (2008 Berg paper reference). Reconstructed: 5d avg TRIN <=0.70 near highs (euphoric) = WARNING; >=2.00 near lows (capitulation) = BUY. USI:TRIN 1971→.

52-week-high count as sentiment/breadth input (named in Berg's indicator list); needs issue-level data.

• Required series not freely available; registered with missing_data.

S&P closes at a fresh 60-day low — base condition of the 1957/1962/2001/2018/2025 bottom models (combined with volume regime).

S&P 500 3-day ROC <= -10% — extreme short-term decline (thresholds list).

S&P 500 5-day ROC <= -10% — extreme weekly decline (thresholds list).

S&P 5-day performance weakest in 180 days — Feb 5 2026 panic marker (only analog 1997, +39% followed).

S&P 5-day performance weakest in 5 years — deeper rarity variant (thresholds list).

S&P constituent up-volume 70x down-volume (Mar 10 2009). Constituents not free; proxy = index up-day on >=2x 20-day volume after >=20% decline.

Component of the 6-occurrence thrust model. Constituent breadth not free; defined with citation, fires when breadth series is ingested.

• Computed series unavailable; cited historical dates are retained.

7-day window containing a 12:1 upside (or downside) volume day with high average volume — 17 occurrences, bull market every instance.

NYSE upside/downside volume >=10:1 three consecutive days — 2 occurrences (1984-08-03, 2010-09-03), +20.5%/+23.4% within a year.

Single day with NYSE upside volume 15x downside — Mar 10 2009 GFC model.

New 60-day low followed by >=+4% within 5 days without retest — the four cited V-bottoms (Mar 9 2009, Dec 24 2018, Mar 23 2020, Apr 8 2025).

S&P-side mirror of the cited VXN optimism-warning (same principle applied to VIX).

VIX gains >=10% for 2 consecutive days — panic acceleration; part of the Apr 4 2025 triple that marked the low.

5-day avg volume highest in 375 days within last 20 days AND price within 7% of 60-day low — the high-volume regime into lows (1957, 1962, 2001, 2018, 2025).

Same volume regime but with price within 5% of 252-day high — bearish volatility warning (S&P 1987, gold/silver 2026).

NASDAQ VIX falling unusually fast vs its 30-day average = optimism extreme; warning (top context). Threshold reconstructed from the rarity principle.

VXN (NASDAQ-100 VIX) gains >=10% for 2 consecutive days — NASDAQ-side panic.

VXN +35% in 3 days with NASDAQ A/D down 8 of 10 and weakest 5-day in 180 — one historical precedent (1997) followed by +39% without pullback.

Rare SELL: VXN short-term ratio vs longer-term at extreme while market near highs. Fired 2025-12-11 (stop +2.75%); 2021 analog +0.94% at day 21 then -21.4%.

Spot VIX minus VIX3M > 0, the best-source-fidelity public variant from qm-4bx.15. The family was rejected as a standalone sleeve because it reduces drawdown at a material CAGR cost; it is published as a weak, independently researched Other indicator.

• VIX3M native daily history begins 2006-07-17; synthetic prehistory is not used by this signal payload.
• qm-4bx.15 rejected promotion: candidate CAGR trails SPY despite lower volatility.

Pine-verbatim VIX/VIX3M ratio EMA50 above EMA252 risk-off regime from qm-kdnj.3. The exact re-audit is rejected economically but remains published as a negative Other indicator with fidelity and warmup disclosure.

• VIX3M native daily history begins 2006-07-17; pre-warmup values are unavailable.
• qm-kdnj.3 rejected promotion after Pine fidelity was fixed and economics remained negative.

SMA50 of INDEX:HIGN minus INDEX:LOWN below zero risk-off regime, the source-baseline variant from qm-4bx.16. The sleeve is rejected on return and Sharpe gates but remains discoverable as a weak breadth indicator.

• Native NYSE high/low breadth begins 2001-06-05.
• Synthetic sector-ETF history is a disclosed NO-GO proxy and is not used for native publication.
• qm-4bx.16 rejected promotion: drawdown improves but CAGR and Sharpe trail SPY.

Pine-verbatim cumulative NH-NL z-score below zero regime from qm-kdnj.4. Fidelity is fixed and defensive timing remains visible, but the corrected result is rejected economically; this negative Other indicator is published without retuning.

• Native HIGN/LOWN history begins 2001-06-05; the corrected Pine panel has a 50-bar warmup.
• No synthetic pre-2005 extension or threshold retuning is used.

Public-proxy reconstruction of the Berg price/turning-boundary family: SPY rolling 20-day percentile <=5%. qm-tykd.1 found the broader 252-day <=5% proxy negative and regime-dependent; this best-source-fidelity proxy is published weakly, not as Berg-verbatim logic.

• Berg disclosed only the family concept; all proxy thresholds and windows are inferred.
• The proprietary 30,000-indicator feed is unavailable and no native Berg data is redistributed.
• qm-tykd.1 rejected the proxy on subperiod, adjacent-variant, and overlay gates.

Public-proxy reconstruction of the Berg volume/TRIN/capitulation family: DVOL/UVOL >1.5. qm-tykd.2’s return-shock proxy was negative across regimes; this source-fidelity volume proxy is published weakly with share-volume and native-feed gaps.

• Berg disclosed only the family concept; threshold and holding window are inferred proxy choices.
• Licensed share-volume/TRIN inputs and the proprietary 30,000-indicator feed are unavailable.
• qm-tykd.2 rejected the tested family proxy on regime, adjacent, and overlay gates.

Public-proxy reconstruction of the Berg breadth/multi-index thrust family: 10-day SMA of HIGN/(HIGN+LOWN) >0.615. All numeric thresholds are inferred; the negative qm-tykd.3 family result remains visible in the card disclosure.

• Berg disclosed only the family concept; 10-day window and 0.615 threshold are inferred proxy choices.
• Native multi-index/proprietary breadth feeds are unavailable; HIGN/LOWN is a public proxy.
• qm-tykd.3 rejected the family on subperiod, adjacent-variant, and overlay gates.

Public-proxy reconstruction of the Berg composite family: at least 1.5 public extreme votes from price and volume proxy components. qm-tykd.4’s richer count>=2 composite was negative versus SPY; this simplest campaign proxy is published with a prominent no-native-rule disclosure.

• Berg disclosed only the composite family; component selection, threshold, and window are inferred.
• The proprietary 30,000-indicator feed, composite weights, and native inputs are unavailable.
• qm-tykd.4 rejected the tested composite on subperiod and overlay gates.