Research evidence and economic schedules
Review dated model diagnostics and original schedule sources. No qualified longer-horizon finding is presented.
Recorded daily diagnostic
Issue dates: 2026-08-31 to 2026-09-29.
3257 of 6638 graded forecasts matched direction (49.1%).
6651 recorded; 6638 graded; 13 ungraded.
Source: stored model_predictions; latest per ticker and New York issue date. Records can be revised.
Recorded daily direction labels are not portfolio returns. Direction accuracy does not establish weekly or monthly buy-and-hold performance.
Four-week follow-up — separate data snapshot, not qualified
August 31–September 28, 2026: research returned -2.82%, versus SPY +0.18% and random accounts averaging -4.95%, after 0.10% per purchase/sale. It lost money and trailed SPY; beating random baskets does not establish an edge. Monthly results remain pending until the October 1 endpoint is measured. This is not this edition's return.
The provider changed 335 historical rows, including 115 with raw-price changes. The strict extension therefore withholds returns. The separately planned calculation below uses the complete new data snapshot and the same frozen baskets. It does not replace the earlier report. Earlier-window net-return differences are 0.00 percentage points across all 18 comparisons.
Scroll the table horizontally on narrow screens to see every comparison.
| Account | 0% | 0.10% | 0.25% |
|---|---|---|---|
| Research baskets | -2.04% | -2.82% | -3.98% |
| SPY buy-and-hold | +0.38% | +0.18% | -0.12% |
| Fixed-universe buy-and-hold | -3.82% | -4.01% | -4.30% |
| SPY with research allocation/rebalancing | +0.38% | -0.42% | -1.61% |
| Fixed universe with research allocation/rebalancing | -3.83% | -4.59% | -5.73% |
| 100 random funded accounts, mean | -4.18% | -4.95% | -6.08% |
The $10,000 research account ends at $9,717.59: $9,711.27 cash plus $6.32 unpaid dividends. Costs total $78.40; drawdown at observed marks is -3.89%. All books now include final sale costs. No actual fills or subscriber performance are claimed.
Cutoff: 2026-09-28 20:00 UTC; observed: 2026-09-29 01:53 UTC. This fixed snapshot does not mature automatically. Both plans are post-inspection; the revised population excludes three known identity conflicts. Publication, original inputs and complete eligibility are unverified. Four periods are insufficient to establish durable skill. Controls are not sector- or beta-matched; related random draws are not independent evidence. Tariff, rate, inflation and dip timing were not tested.
Both complete twelve-case records, including rejected and pending results. Each line covers 0%, 0.10% and 0.25% costs per side:
- Strict extension; Original (238 names); weekly; all three costs: incomplete; 4/4 periods matured, 0 pending; 3 known identity conflicts.
- Strict extension; Original (238 names); monthly; all three costs: pending; 0/1 periods matured, 1 pending; 3 known identity conflicts.
- Strict extension; Revised (235 names); weekly; all three costs: incomplete; 4/4 periods matured, 0 pending; 0 known identity conflicts.
- Strict extension; Revised (235 names); monthly; all three costs: pending; 0/1 periods matured, 1 pending; 0 known identity conflicts.
- Separate new data snapshot; Original (238 names); weekly; all three costs: incomplete; 4/4 periods matured, 0 pending; 3 known identity conflicts.
- Separate new data snapshot; Original (238 names); monthly; all three costs: pending; 0/1 periods matured, 1 pending; 3 known identity conflicts.
- Separate new data snapshot; Revised (235 names); weekly; all three costs: measured; not qualified; 4/4 periods matured, 0 pending; 0 known identity conflicts.
- Separate new data snapshot; Revised (235 names); monthly; all three costs: pending; 0/1 periods matured, 1 pending; 0 known identity conflicts.
Funded weekly/monthly study — exploratory, hypothetical, not qualified
The revised weekly research account returned -1.79% after 0.10% costs per purchase and sale, versus SPY buy-and-hold +0.01%. This short study does not establish a profitable or predictive edge. Monthly results remain pending.
Measured window: 2026-08-31 13:30 UTC to 2026-09-21 13:30 UTC; 3 of 4 weekly periods completed. Fixed outcome cutoff: 2026-09-25 20:00 UTC; evidence observed 2026-09-28 00:17 UTC. Pending cases do not update automatically. This is a historical appendix, not this edition's weekly return.
Each account starts with $10,000.00 once, with no later contributions, borrowing or shorting. Research selects up to ten bullish daily forecasts by saved score within a fixed population, equally weighted, entering at the next session open and holding to the next calendar boundary. Scores are not calibrated probabilities. These are not weekly/monthly model forecasts or actual subscriber returns.
Scroll the table horizontally on narrow screens to see every comparison.
| Account | 0% | 0.10% | 0.25% |
|---|---|---|---|
| Research baskets | -1.20% | -1.79% | -2.67% |
| SPY buy-and-hold | +0.11% | +0.01% | -0.14% |
| Fixed-universe buy-and-hold | -2.59% | -2.69% | -2.84% |
| SPY with research allocation/rebalancing | +0.11% | -0.49% | -1.38% |
| Fixed universe with research allocation/rebalancing | -2.55% | -3.13% | -4.00% |
| 100 random funded accounts, mean | -2.91% | -3.49% | -4.36% |
Ending research value at primary costs: $9,821.21, comprising $9,811.16 spendable cash and $10.05 unpaid dividend receivables. Modeled costs: $58.88; maximum drawdown at observed marks: -3.89%, not every intraday trough. Splits and dividend payment timing are included; cash earns zero interest.
The partial valuation follows the September 21 research sale and precedes its next purchase. Buy-and-hold controls remain invested and have not paid a final exit cost. September 21–25 performance is excluded. Allocation-matched controls use the research trading schedule. Random accounts use the same period and costs; their mean is not a probability of success. Controls are not sector- or beta-matched.
All twelve cases, including incomplete and pending results:
- Original, 238 names; weekly; 10 bps per side: incomplete; 3/4 periods matured, 1 pending; 3 known identity conflicts.
- Original, 238 names; weekly; 0 bps per side: incomplete; 3/4 periods matured, 1 pending; 3 known identity conflicts.
- Original, 238 names; weekly; 25 bps per side: incomplete; 3/4 periods matured, 1 pending; 3 known identity conflicts.
- Original, 238 names; monthly; 10 bps per side: pending; 0/1 periods matured, 1 pending; 3 known identity conflicts.
- Original, 238 names; monthly; 0 bps per side: pending; 0/1 periods matured, 1 pending; 3 known identity conflicts.
- Original, 238 names; monthly; 25 bps per side: pending; 0/1 periods matured, 1 pending; 3 known identity conflicts.
- Revised, 235 names; weekly; 10 bps per side: partial; 3/4 periods matured, 1 pending; 0 known identity conflicts.
- Revised, 235 names; weekly; 0 bps per side: partial; 3/4 periods matured, 1 pending; 0 known identity conflicts.
- Revised, 235 names; weekly; 25 bps per side: partial; 3/4 periods matured, 1 pending; 0 known identity conflicts.
- Revised, 235 names; monthly; 10 bps per side: pending; 0/1 periods matured, 1 pending; 0 known identity conflicts.
- Revised, 235 names; monthly; 0 bps per side: pending; 0/1 periods matured, 1 pending; 0 known identity conflicts.
- Revised, 235 names; monthly; 25 bps per side: pending; 0/1 periods matured, 1 pending; 0 known identity conflicts.
The revised population excludes three documented identity conflicts from the original 238 names. Original cases remain visible; this does not certify eligibility of the other 235 names. The plan was fixed before these funded calculations but after viewing earlier cohort results and identity findings. This post-inspection amendment, small sample and related observations prevent a skill claim. Original publication timing, input availability and complete historical eligibility remain unverified. Longer holding and dip timing still require separate prospective tests.
Separate SPY study: longer holds did not establish an edge
December 31, 2024–December 16, 2025. All 16 model/holding combinations lagged SPY buy-and-hold (+15.81%) after 0.10% per purchase/sale. Model ranges: 5 sessions: +2.34% to +4.86%; 10 sessions: +5.71% to +8.83%; 20 sessions: +8.47% to +13.11%; 60 sessions: +12.50% to +15.16%. None improved measured drawdown.
Each account starts with $10,000 once; no borrowing or later deposits. Four saved weekly/monthly model variants choose SPY when the preceding session's score exceeds 0.5, otherwise cash. Scores are not established probabilities of profit. A/B use different historical VWAP target formulas (typical-price/midpoint), not quality grades. Session holds approximate weeks/months; they are not calendar-month portfolios. Tables scroll horizontally on narrow screens.
| Model / sessions | Model return | Always SPY, same trades | Random timing mean | Model drawdown |
|---|---|---|---|---|
| Weekly A / 5 | +10.41% | +16.13% | +13.91% | 19.91% |
| Weekly A / 10 | +9.58% | +16.11% | +13.31% | 19.91% |
| Weekly A / 20 | +10.66% | +16.11% | +13.60% | 19.91% |
| Weekly A / 60 | +13.18% | +16.08% | +10.98% | 19.91% |
| Monthly A / 5 | +11.40% | +16.13% | +11.89% | 19.91% |
| Monthly A / 10 | +13.27% | +16.11% | +14.54% | 19.91% |
| Monthly A / 20 | +15.62% | +16.11% | +14.50% | 19.91% |
| Monthly A / 60 | +16.08% | +16.08% | +16.08% | 19.91% |
| Weekly B / 5 | +14.03% | +16.13% | +15.43% | 19.91% |
| Weekly B / 10 | +12.91% | +16.11% | +14.80% | 19.91% |
| Weekly B / 20 | +15.62% | +16.11% | +15.34% | 19.91% |
| Weekly B / 60 | +16.08% | +16.08% | +16.08% | 19.91% |
| Monthly B / 5 | +10.09% | +16.13% | +15.45% | 19.91% |
| Monthly B / 10 | +12.61% | +16.11% | +13.91% | 19.91% |
| Monthly B / 20 | +15.62% | +16.11% | +14.35% | 19.91% |
| Monthly B / 60 | +16.08% | +16.08% | +16.08% | 19.91% |
| Model / sessions | Model return | Always SPY, same trades | Random timing mean | Model drawdown |
|---|---|---|---|---|
| Weekly A / 5 | +2.34% | +5.51% | +5.58% | 21.25% |
| Weekly A / 10 | +5.71% | +10.67% | +9.31% | 20.39% |
| Weekly A / 20 | +8.47% | +13.36% | +11.36% | 20.23% |
| Weekly A / 60 | +12.50% | +15.16% | +10.32% | 20.07% |
| Monthly A / 5 | +3.87% | +5.51% | +4.34% | 21.25% |
| Monthly A / 10 | +8.83% | +10.67% | +10.05% | 20.39% |
| Monthly A / 20 | +13.11% | +13.36% | +12.01% | 20.23% |
| Monthly A / 60 | +15.16% | +15.16% | +15.16% | 20.07% |
| Weekly B / 5 | +4.86% | +5.51% | +6.14% | 21.25% |
| Weekly B / 10 | +8.49% | +10.67% | +10.30% | 20.39% |
| Weekly B / 20 | +13.11% | +13.36% | +12.83% | 20.23% |
| Weekly B / 60 | +15.16% | +15.16% | +15.16% | 20.07% |
| Monthly B / 5 | +2.45% | +5.51% | +7.44% | 21.25% |
| Monthly B / 10 | +8.42% | +10.67% | +9.67% | 20.39% |
| Monthly B / 20 | +13.11% | +13.36% | +11.87% | 20.23% |
| Monthly B / 60 | +15.16% | +15.16% | +15.16% | 20.07% |
| Model / sessions | Model return | Always SPY, same trades | Random timing mean | Model drawdown |
|---|---|---|---|---|
| Weekly A / 5 | -8.68% | -8.63% | -5.78% | 23.81% |
| Weekly A / 10 | +0.15% | +2.99% | +3.57% | 21.49% |
| Weekly A / 20 | +5.27% | +9.36% | +8.07% | 20.71% |
| Weekly A / 60 | +11.50% | +13.79% | +9.33% | 20.31% |
| Monthly A / 5 | -6.48% | -8.63% | -6.06% | 23.81% |
| Monthly A / 10 | +2.50% | +2.99% | +3.65% | 21.49% |
| Monthly A / 20 | +9.45% | +9.36% | +8.38% | 20.71% |
| Monthly A / 60 | +13.79% | +13.79% | +13.79% | 20.31% |
| Weekly B / 5 | -7.55% | -8.63% | -6.41% | 23.81% |
| Weekly B / 10 | +2.18% | +2.99% | +3.89% | 21.49% |
| Weekly B / 20 | +9.45% | +9.36% | +9.17% | 20.71% |
| Weekly B / 60 | +13.79% | +13.79% | +13.79% | 20.31% |
| Monthly B / 5 | -8.02% | -8.63% | -3.55% | 23.81% |
| Monthly B / 10 | +2.42% | +2.99% | +3.60% | 21.49% |
| Monthly B / 20 | +9.45% | +9.36% | +8.24% | 20.71% |
| Monthly B / 60 | +13.79% | +13.79% | +13.79% | 20.31% |
Source: retained EODHD raw daily prices and dividend/payment records observed September 30, 2026 UTC; retrospectively fitted SPY classifiers. These are hypothetical fractional-share references, not fills or subscriber returns. At 0.10% costs, SPY buy-and-hold ends at $11,580.77 with $21.49 costs and 19.91% marked drawdown; five-session model costs range about $690–$830.
Every scheduled boundary sells and repurchases, even if SPY remains selected. Actual position netting was not tested in this original study. Cash earns no interest; dividends become spendable on payment date. Buy-and-hold retains dividends as cash; scheduled accounts can reinvest them. This explains the small zero-cost advantage of always-long sixty-session cases, not selection skill. Drawdown uses open/close and trading marks, not every intraday trough.
All 48 cases resolved. Score coverage is 46/48, 23/24, 12/12 and 4/4 slots for 5/10/20/60 sessions; missing scores stay cash. Each case includes 100 frozen random timing draws preserving invested-block counts and missing slots. Draws repeat and share one market path, so they are not independent evidence or probabilities of success. Three sixty-session variants always hold SPY: their random controls are identical.
Post-inspection research, not prospective validation. The original calendar omitted the January 9, 2025 market closure; the corrected schedule was frozen after initial prices were observed but before returns were calculated. Original records remain retained. Historical feature availability and publication timing remain unverified; current vendor history is not a historical-vintage archive. The underlying classifier labels overlap. No durable predictive skill is established. Tariff, rate, inflation and dip timing were not separately tested. This is historical SPY timing research, not the earlier stock-basket study, this edition's return, or a current trade list. Results remain not qualified.
Follow-up: timing skill remains unproven
Retaining unchanged positions improved the original study, but none of those models beat buy-and-hold. Separate models trained on actual holding-period outcomes had mixed results. One favorable monthly result depended on a single cash interval; a later check did not establish repeatable timing skill.
Later funded window: February 27–June 9, 2026, 70 session intervals. Each account starts with $10,000, holds SPY or cash, retains unchanged shares, pays dividends into cash and assumes zero cash interest and 0.10% per purchase/sale. SPY buy-and-hold returned +8.91%, with 7.67% measured maximum drawdown.
| Hold | Net return | Maximum drawdown | Random timing mean | Invested blocks |
|---|---|---|---|---|
| 5 sessions | +7.65% | 4.28% | +2.74% | 5/14 |
| 10 sessions | +8.91% | 7.67% | +8.91% | 7/7 |
| 20 sessions | +8.91% | 7.67% | +8.91% | 4/4 |
| 21 sessions | +8.91% | 7.67% | +8.91% | 4/4 |
The 5-session model earned less than buy-and-hold but had a smaller measured drawdown. The 10-, 20- and 21-session models stayed invested throughout, so their returns and random timing controls equal buy-and-hold. No additional timing skill is demonstrated by those ties. Random timing means use 100 saved permutations per horizon on the same price path; they are not independent trials or confidence levels. The final 20- and 21-session blocks contain only 10 and 7 sessions. Matching invested-block counts does not match invested days for these partial blocks. Drawdowns use sampled prices, not intraday extremes.
Earlier funded window: December 31, 2024–December 16, 2025, $10,000 once, SPY or cash, unchanged shares retained, 0.10% per purchase/sale. SPY buy-and-hold returned +15.81%. Later direction check: 71 signal dates, February 26–June 8, 2026, using frozen models without retraining. Later accuracy is not a portfolio return.
| Hold | Earlier return | Later accuracy | Always-positive accuracy | Later positive forecasts |
|---|---|---|---|---|
| 5 sessions | +15.86% | 43.66% | 61.97% | 28/71 |
| 10 sessions | +16.26% | 59.15% | 59.15% | 71/71 |
| 20 sessions | +27.38% | 69.01% | 69.01% | 71/71 |
| 21 sessions | +14.31% | 71.83% | 71.83% | 71/71 |
The 20-session advantage came from one uninterrupted cash period, January 31–March 31, 2025. The 10-session gain came from missing forecasts forcing cash, not negative model signals. The 5-session advantage disappeared at higher assumed costs (0.25% each side). The 21-session schedule ends with a shortened nine-session block. The 60-session model failed a training/evaluation class requirement and remains excluded from performance claims.
Hypothetical raw-open prices and assumed costs are not fills or subscriber returns. Historical feature availability remains unverified; overlapping labels and repeatedly inspected history limit the evidence. No tariff, rate or inflation forecasting skill is established. This is SPY timing research, not the full emailed stock portfolio or a current trade list. Results remain not qualified.
Economic schedule and sources
Economic calendar — 2026-10-01Selected high-importance US economic and policy events.
No matching events in the reviewed coverage for this date.
US release schedule pins through 2026-12-31; selected releases only.
Importance is MoneySignals editorial classification. These are schedule facts; actual, forecast and previous values are not supplied. Source review dates are not live update times. Schedules can change.
Research over weeks and months
Holding-period research must freeze selections before entry and compare returns after costs and dividends with the same-window benchmark, retaining losses.
No qualified longer-horizon finding is available in this edition. A scheduled release does not establish the direction of the market response.