quantmodel

← Results index

Replicate RecessionAlert indicator: Bear Market High-Risk Alert | RecessionAlert

2026-09-01T00:00:00Z Experiment completed The SHA-verified RecessionAlert Reflections article discloses CMHI Diffusion-zero cash advice after 3 of 7 bearish paw prints, but no native CMHI series, complete formulas, publication lag, or BUY/re-entry map. A best-effort public-proxy campaign executed seven labeled variants over the longest checked-in history: the causal 12-month one-bar-lag proxy returned 9.36% annualized with a -24.66% drawdown, while the lag-0 lookahead control returned 19.84%; results are exploratory, unpromoted, and published with source-fidelity gaps.
1359 recessionalert research_campaign cmhi bear-market high-risk-alert diffusion data_limited fidelity_mismatch parameter_sensitive needs_more_research

Source report: research/findings/1359.md

Objective

Independently replicate the distinct RecessionAlert row Bear Market High-Risk Alert | RecessionAlert from the checked-in canonical snapshot at data/recessionalert.sqlite, without live-site requests, credentials, private account data, or mutation of the crawl database. Because the source is incomplete, this run also performs the required labeled best-effort public-proxy exploration instead of stopping at a gap-only result.

Approach

The canonical page is bound across research/findings/recessionalert_inventory.md, research/artifacts/recessionalert/manifest.json, and read-only data/recessionalert.sqlite to SHA256 sha256:9bead77031d36f389d1b2204111f926fce443af211023b6ccda3a4febb59000a. The selected HTML body is 93,127 bytes, HTTP 200, fetched at 2026-08-22T23:45:17Z; the article publishes at 2015-08-25T17:23:00-0400. The snapshot fetch interval is archival delay, not CMHI publication lag. The target page has zero canonical assets, and no cached HTML/PDF/chart bytes are redistributed.

The campaign uses read-only data/market.sqlite observations: SPY, checked-in synthetic SPX breadth-200/breadth-50 neighbors, Yahoo ^VIX, FRED INDPRO and UNRATE substitutions for unavailable CFNAI/VIXCLS/weekly-leading candidates, and IEF/BIL inferred defensive sleeves. The monthly panel spans 1993-01-31 through 2026-07-31; the IEF-overlap portfolio window begins 2002-08-31. No live refresh or database import was performed.

Disclosed rule and interpretation

The source says that market-internals deterioration over the prior 20 trading days, especially the 2015-08-18 through 2015-08-24 plunge, put CMHI Stock Market Health Diffusion at zero, advising zero stock-market exposure / cash positions. It says 3 of 7 bearish Bear-market paw prints are sufficient for Diffusion zero, describes below-zero new-high breadth, above-long-term-MA breadth, and advancing-volume breadth, and forecasts a Seasonality +1 → −1 flip on 2015-09-01. Seasonality, Momentum-1, Momentum-2, and Weekly Economic Index are named as still above water at the article date.

No BUY/re-entry rule, complete seven-component construction, instrument/weight map, holding period, fill/cost policy, same-day timing, or measured CMHI release lag is disclosed. Those omissions remain explicit source-fidelity gaps; proxy metrics below are not native CMHI evidence.

Causal policy

Treat article information as unavailable before 2015-08-25T17:23:00-0400. Align proxy inputs at month end using latest available observations without future backfill. Execute only after the next available bar plus at least one available-bar lag; lag two is a sensitivity. Unknown observations stay unavailable rather than false, neutral, imputed, or carried before inception. Lag zero is shown only as a lookahead control and is not admissible evidence.

Executed best-effort variants

All rows use the same IEF-overlap window unless stated otherwise. Returns are descriptive close-only overlay metrics, not investment advice or a promotion gate.

VariantLagRisk-off monthsEntriesTotal returnAnnualizedMax drawdown
12-month 3-of-7 best guess1167/288407.48929.36%-24.66%
6-month 3-of-7 neighbor1156/288513.93196.90%-22.95%
21-month, -0.5 threshold neighbor1128/288307.61609.42%-27.17%
12-month 3-of-7 extra-lag neighbor2166/288396.40088.73%-19.08%
Market-internals family1124/288446.89249.02%-26.44%
Macro-confirmed family (current duplicate)1167/288407.48929.36%-24.66%
Lag-0 lookahead control0168/2884074.817119.84%-16.88%

The SPY context benchmark is 10.78% annualized over 1993-01–2026-07; IEF is 3.51% from 2002-07; BIL is 1.38% from 2007-05. The BIL best-guess sensitivity is 7.32% annualized from 2007-06–2026-07, but sleeve spans differ. The best guess is risk-off at proxy scores 6 and 7 around the 2015 source analog, remains defensive during the 2008-09 and 2020 selloff analogs, and loses 17.90% in 2022 while risk-off for 11 of 12 months.

Lag sensitivity is decisive: the lookahead control's 19.84% annualized result is much higher than the causal 9.36%/8.73% results. The 21-month strict neighbor is nearly flat in 2022-01–2026-07 with 0.05% annualized return and -23.12% drawdown. Macro-confirmed warnings agree exactly with three-of-seven under the current implementation; market-only warnings disagree in 43 months. These are retained negative/duplication findings, not hidden or tuned away.

Explicit gaps and disposition

Disposition: exploratory, unpromoted. Publish the disclosed rule and the executed proxy campaign under Other / Research, with tested signal/equity charts and all limitations. Do not register an official CMHI indicator or claim a source-faithful BUY/SELL signal. Reconsider after an authorized checked-in artifact provides native series, formulas, release metadata, and execution semantics.

Files and reproducibility

Source artifact: research/results/1359.md