Research campaign · Iteration 1 · unassessed
RecessionAlert TRENDEX SP-500 — exploratory proxy campaign
Can the incomplete RecessionAlert TRENDEX concept yield transparent causal exploratory proxy signals from the longest defensible offline public history without claiming native source fidelity?
- Expressions
- 5
- Logged trials
- Not recorded
- Independent events
- Not assessed
- Evidence
- unassessed
What the research found
The source-backed result is an incomplete support/resistance trend-duration and historical-reversal-probability concept; exact native observations, formula, STOP calculation, threshold, release lag, action mapping, and portfolio semantics are absent.
Best-effort exploration uses normalized checked-in SPY history from 1993-01-29 through 2026-07-24 and a SPY/IEF common overlay from 2002-07-31 through 2026-07-24. IEF/BIL are public defensive proxies only.
Four causal variants and one same-close sensitivity were executed. The two-bar breakout analogue has the highest causal Sharpe but is likely generic trend timing; the channel and SMA variants trade return for drawdown.
Same-close behavior is materially different and non-causal. All outputs are exploratory proxy evidence and do not register a native TRENDEX indicator or claim RecessionAlert fidelity.
Mechanism and falsifiers
Not recorded in this iteration.
Not recorded in this iteration.
Confidence and limitations
Native TRENDEX support/resistance values, trend counts, probabilities, STOP levels, source-linked series, and numeric native span are unavailable.
Exact support/resistance construction, reversal definition, empirical calibration, score range, threshold, persistence, state machine, and top/bottom action mapping are undisclosed.
No recurring calculation/release clock, timezone, information cutoff, point-in-time vintage, or revision policy is disclosed. The article p(n+1) horizon is not a release lag.
SPY is an investable ETF proxy for the S&P context; IEF/BIL are defensive overlay choices. None is a claimed TRENDEX input or output.
Overlay metrics are gross descriptive results on a common window and do not include costs, slippage, tax, or a native portfolio rule. Same-close results are non-causal and excluded from candidate claims.
The campaign does not merge Demark/Demark+, SP500 Probability Model, SP500 TOP composite, Standard Charts, adjacent workbooks, or uncaptured PRO media into this distinct source identity.
Not recorded in this iteration.
Compare expressions
Download evidenceExploratory results. Check each period, proxy and cost assumption before comparing. — means not recorded.
| Expression / family | CAGR | Sharpe | Max drawdown | Test period | Assessment |
|---|---|---|---|---|---|
| trendex_proxy_channel_21_lag1channel-rank | — | 0.51 | -41.7% | Not recorded |
exploratory |
| trendex_proxy_channel_50_lag1channel-rank | — | 0.55 | -46.3% | Not recorded |
exploratory |
| trendex_proxy_sma_50_lag1sma-distance | — | 0.59 | -40.5% | Not recorded |
exploratory |
| trendex_proxy_breakout_20_lag2breakout-stop | — | 0.67 | -53.5% | Not recorded |
exploratory candidate |
| trendex_proxy_channel_same_closechannel-rank | — | -1.91 | -100.0% | Not recorded |
research only |
Open an expression to inspect its rules and request confirmation. The request must be submitted by a trusted repository collaborator.
trendex_proxy_channel_21_lag1
- created by
inferred_best_guess
- family id
channel-rank
- parameters
- holding period
1
- lag bars
1
- lookback
21
- threshold
0.6
- reason tested
Closest public support/resistance rank analogue with causal one-bar lag.
- status
exploratory
- tags
data_limited
fidelity_mismatch
useful_as_portfolio_overlay
- variant id
trendex_proxy_channel_21_lag1
trendex_proxy_channel_50_lag1
- created by
inferred_nearby_variant
- family id
channel-rank
- parameters
- holding period
1
- lag bars
1
- lookback
50
- threshold
0.7
- reason tested
Nearby longer channel and higher threshold.
- status
exploratory
- tags
parameter_sensitive
fidelity_mismatch
- variant id
trendex_proxy_channel_50_lag1
trendex_proxy_sma_50_lag1
- created by
inferred_nearby_variant
- family id
sma-distance
- parameters
- holding period
1
- lag bars
1
- lookback
50
- threshold
0.7
- reason tested
Nearby moving-average contrast despite the source preference for support/resistance.
- status
exploratory
- tags
interesting_negative_result
fidelity_mismatch
- variant id
trendex_proxy_sma_50_lag1
trendex_proxy_breakout_20_lag2
- created by
inferred_speculative_extension
- family id
breakout-stop
- parameters
- holding period
1
- lag bars
2
- lookback
20
- threshold
0.5
- reason tested
Prior-channel breakout and conservative two-bar lag as a black-line/STOP analogue.
- status
exploratory_candidate
- tags
interesting_negative_result
candidate_for_confirmation
- variant id
trendex_proxy_breakout_20_lag2
trendex_proxy_channel_same_close
- created by
inferred_timing_sensitivity
- family id
channel-rank
- parameters
- holding period
1
- lag bars
0
- lookback
21
- threshold
0.6
- reason tested
Explicit same-close sensitivity; never a causal baseline.
- status
research_only
- tags
interesting_negative_result
fidelity_mismatch
needs_more_research
- variant id
trendex_proxy_channel_same_close
Interactive lab
8428 observationsLoading available evidence…
Research record
Source claims, inferred rules, experiments and the evidence behind the assessment.
Source
- content length
80141
- database path
data/recessionalert.sqlite
- fetched at
2026-08-22T23:46:43Z
- inventory path
research/findings/recessionalert_inventory.md
- kind
RecessionAlert inventory row
- manifest path
research/artifacts/recessionalert/manifest.json
- native span
One dated narrative article; no source-linked TRENDEX numeric observation span
- publication lag
Article-to-snapshot interval is archival delay (2544.372604166667 days), not model release lag; recurring release timestamp is undisclosed
- published at
2019-09-04T10:50:10-0400
- sha256
sha256:0847fa3d8bd7a9dc74a9e07e000ec338362ad84c743b57d710fab8f5d4b8119c
- title
About the new TRENDEX SP-500 model | RecessionAlert
What the source claims
The article says TRENDEX supersedes Demark and Demark+ trend counting and focuses on support and resistance rather than arbitrary closes four days ago and moving averages.
The article says trend-duration counts are dynamically mapped to historical successful-reversal probabilities, with probabilities of market tops and market bottoms and a role in the SP500 TOP model.
The chart legend defines STOP as the current-trade cancel level, p(n) as the last displayed probability, and p(n+1) as tomorrow's probability if price does not cross the black line.
The page does not disclose source-linked observations, support/resistance construction, trend-count initialization, reversal calibration, thresholds, state transitions, recurring point-in-time release, or a portfolio action rule.
A public SPY rolling channel-rank score is used as the closest transparent support/resistance analogue, with one available-bar execution lag and IEF as an exploratory defensive sleeve.
A public SMA-distance rank is retained as a nearby contrast, while a prior-channel breakout is a speculative black-line/STOP analogue; neither is native TRENDEX.
Five proxy variants, including four causal lagged variants and one same-close sensitivity, are tested on the longest checked-in SPY history and common SPY/IEF overlay window.
Rules actually disclosed
The article says TRENDEX supersedes Demark and Demark+ trend counting and focuses on support and resistance rather than arbitrary closes four days ago and moving averages.
The article says trend-duration counts are dynamically mapped to historical successful-reversal probabilities, with probabilities of market tops and market bottoms and a role in the SP500 TOP model.
The chart legend defines STOP as the current-trade cancel level, p(n) as the last displayed probability, and p(n+1) as tomorrow's probability if price does not cross the black line.
The page does not disclose source-linked observations, support/resistance construction, trend-count initialization, reversal calibration, thresholds, state transitions, recurring point-in-time release, or a portfolio action rule.
What had to be inferred
A public SPY rolling channel-rank score is used as the closest transparent support/resistance analogue, with one available-bar execution lag and IEF as an exploratory defensive sleeve.
A public SMA-distance rank is retained as a nearby contrast, while a prior-channel breakout is a speculative black-line/STOP analogue; neither is native TRENDEX.
Five proxy variants, including four causal lagged variants and one same-close sensitivity, are tested on the longest checked-in SPY history and common SPY/IEF overlay window.
Research questions
Can the incomplete RecessionAlert TRENDEX concept yield transparent causal exploratory proxy signals from the longest defensible offline public history without claiming native source fidelity?
Data
- canonical sha256
sha256:0847fa3d8bd7a9dc74a9e07e000ec338362ad84c743b57d710fab8f5d4b8119c
- canonical snapshot
data/recessionalert.sqlite pages row for https://recessionalert.com/about-the-new-trendex-sp-500-model/
- native spans
One dated 2019-09-04 narrative article; no source-linked TRENDEX observation span. SPY proxy 1993-01-29/2026-07-24; SPY/IEF overlay 2002-07-31/2026-07-24.
- proxy data
data/market.sqlite read-only SPY/IEF/BIL public snapshots; normalized derived lab series only
- publication lag
Article-to-snapshot interval 2544.372604166667 days is archival delay; model calculation/release lag is unknown. Causal proxy default is one available-bar lag; two-bar sensitivity and same-close warning retained.
- redistribution
Publish normalized derived series, metrics, metadata, and summaries only; no cached RecessionAlert HTML, PRO chart/image bytes, workbook bytes, credentials, cookies, or private account data.
- revision policy
Checked-in snapshots are fixed for this campaign; source page crawl revisions are not TRENDEX data vintages, and proxy revision history is not reconstructed.
Baseline implementation
- created by
inferred_best_guess
- family id
channel-rank
- metrics
- annualized return
0.0741554876
- cumulative return
4.738
- max drawdown
-0.4170714785
- risk off rate
0.5880013258
- sharpe
0.509947723
- parameters
- holding period
1
- lag bars
1
- lookback
21
- threshold
0.6
- reason tested
Closest public support/resistance rank analogue with causal one-bar lag.
- status
exploratory
- tags
data_limited
fidelity_mismatch
useful_as_portfolio_overlay
- variant id
trendex_proxy_channel_21_lag1
What to try interactively
- default
0.6
- name
threshold
- options
0.5
0.6
0.7
- type
float
- default
21
- name
lookback
- options
20
21
50
- type
integer
- default
1
- name
lag_bars
- options
0
1
2
- type
integer
- default
IEF
- name
defensive_sleeve
- options
IEF
BIL
- type
enum
Suggested next research
Can a separately authorized refresh provide source-linked TRENDEX probability and STOP observations with point-in-time timestamps?
What price field, support/resistance estimator, count initialization, reversal horizon, and empirical denominator define the native model?
Can public constituent breadth history improve the conceptual proxy without being presented as native TRENDEX data?
Do the causal proxy candidates survive transaction costs, persistence, alternative defensive sleeves, and an untouched-period protocol?
Trial ledger
- iteration number
1
- objective
offline best-effort proxy exploration, diagnostics, and public synthesis
Regime behavior
Channel-rank proxy is heterogeneous: CAGR is -1.94% in 2020-2022 and 9.03% in 2023-2026. Breakout is 3.66% and 22.30% in those periods respectively.
Entry-event forward SPY returns are positive on average for each tested proxy; this does not validate the missing native reversal-probability calibration.
State-space exploration
Five variants are retained: channel rank 21/lag 1, channel rank 50/lag 1, SMA-distance 50/lag 1, breakout 20/lag 2, and channel same-close sensitivity.
Common-window baseline is SPY buy-and-hold CAGR 11.16%, Sharpe 0.657, max drawdown -55.19%; all proxy metrics are descriptive.
Timing and cost sensitivity
Channel rank 21 has lag-1 CAGR 7.42% and lag-2 CAGR 8.30%; the zero-lag sensitivity collapses to -28.13% CAGR and -99.96% max drawdown.
No cost claim is made; point-in-time availability is the dominant unresolved source constraint.
What the signal looks like
Interactive Lab uses normalized SPY as the price proxy, shows lagged inferred trigger events, and keeps same-close sensitivity explicitly non-causal.
The displayed scores are channel-rank, SMA-distance, and prior-channel breakout analogues; they are not native TRENDEX probabilities or STOP values.
Next questions
Can a separately authorized refresh provide source-linked TRENDEX probability and STOP observations with point-in-time timestamps?
What price field, support/resistance estimator, count initialization, reversal horizon, and empirical denominator define the native model?
Can public constituent breadth history improve the conceptual proxy without being presented as native TRENDEX data?
Do the causal proxy candidates survive transaction costs, persistence, alternative defensive sleeves, and an untouched-period protocol?
Review status
published. Research publication does not imply official admission.
Return to pending research