← Results index

Replicate RecessionAlert indicator: The SP-500 Great Trough Detector Project

2026-09-03T00:00:00Z Research completed Published the fail-closed Great Trough Detector specification from SHA-verified offline fixtures (project page af1b726b plus siblings d909e1ae/c3ab47de/e6cafff2 and WeeklyData.xlsx fc2be5f1): disclosed 27.4/10/80 Class A/B semantics, native and effective spans 2015-03-26 to 2026-08-14, NYSE daily-close cadence with +1 bar execution, revision limitations, and an explicit no-go for reconstruction or strategy admission with the negative baseline retained on the Other-indicators page.
great-trough recessionalert trough-detector fail-closed

Objective

Follow-up to #911 for the captured RecessionAlert indicator page at https://recessionalert.com/the-sp-500-great-trough-project/ pinned to sha256:af1b726bcf6585f49a61e71fbd8ad43060b8a301ff79f378b24c4c75231d64dc. Review only the durable offline snapshot; do not live-crawl or invent thresholds. Acceptance: publish a fail-closed indicator/strategy specification with page and asset SHA256 provenance, disclosed semantics, native and effective spans, timing/cadence, revision limitations, and an explicit no-go disposition where evidence is insufficient.

Approach

  1. Verified the offline snapshot in data/recessionalert.sqlite: project page

af1b726bcf6585f49a61e71fbd8ad43060b8a301ff79f378b24c4c75231d64dc (2012-06-19, 99190 bytes, fetched 2026-08-22T23:39:25Z) plus siblings great-trough-detector-improvements (d909e1ae), great-trough-signal (c3ab47de), on-standby-for-a-great-trough (e6cafff2), and subscriber WeeklyData.xlsx sheet SP500 BREADTH DATA (fc2be5f1fc2a2e38949ae566f6e8cb4699a95562975a498fb5a53510cc40d67c, 8,209,462 bytes). No live fetch at any step.

  1. Extracted disclosed semantics: canonical 2014 improved Class A/B rules

(Class B arm below 27.4 for 5 trading days, Class A arm below 10 for 10 trading days cancelling B, both buying on a punch above 80 after at least 4 days below 80), severity bands 1-3/4-6/>6 days below 10, NYSE daily-close cadence, next-session (+1 bar) execution, no disclosed exit. Legacy 2012 27/87/14-day and 26/87/15-day variants documented but not executed.

  1. Confirmed native and effective spans 2015-03-26 to 2026-08-14 (2,864 NYSE

sessions; 27 Class A, 58 Class B, 83 unique markers) via the offline runner research/recessionalert/great_trough.py (--check-causality --json passed: monotonic daily index, no pre-2015 fill, late-input immutable).

  1. Froze the fail-closed specification in

research/findings/specs/recessionalert-great-trough.json with explicit no-go items: proprietary breadth smoothing constant, point-in-time NYSE constituents/exclusions, pre-2015 machine history (chart back to 1987, no OCR), workbook marker code 1 vs 2 meaning, and revision/vintage policy.

  1. Retained the existing inspection-only publication: playable Other-indicators

entry recessionalert-great-trough, experiment notebook research/experiments/qm-09tp.22.ipynb with self-contained analysis docs/experiments/qm-09tp.22-analysis.html, and the negative forward baseline (combined markers trail unconditional S&P returns at 37/83/119/150 trading sessions). No strategy or excess-return claim is made.

Files / code / data changed

research/findings/specs/recessionalert-great-trough.json, research/fixtures/buy-signals-v1.json, research/experiments/qm-09tp.22.json/.ipynb, docs/site-data/signals/history.json, docs/other-indicators.html, docs/experiments/qm-09tp.22-analysis.html

Validation

Results

Published the fail-closed Great Trough Detector specification with full page and workbook SHA256 provenance, disclosed Class A/B semantics, native and effective spans, timing/cadence, revision limitations, and an explicit no-go for exact reconstruction and strategy admission. The inspection-only Other-indicators replay (83 exact supplied markers) and its negative forward baseline remain publicly visible with browser coverage. Changelog entry research/results/1593.md is searchable via git grep 1593.

Links

Source artifact: research/results/1593.md