TL;DR
The original five-stock dividend portfolio beat matched SPY from January 2025 through September 2026: 41.04% versus 30.96%. It also fell further, with 24.33% maximum drawdown versus SPY’s 19.92%.
The new three-stock finalist returned 27.66%, despite much stronger earlier growth. Reselecting across four historical periods improved overall growth but lost to controls in several periods. These simulations used already known dates; concentration did not deliver the recent improvement the earlier tests suggested.
The published five-stock portfolio still wins the recent test
Portfolio value (USD)
- Published five-stock portfolio 2026-09-30, 16:00:00 EDT
- $14,104.39
- New three-stock finalist 2026-09-30, 16:00:00 EDT
- $12,766.49
- SPY with matching fees and reinvestment 2026-09-30, 16:00:00 EDT
- $13,096.36
Published five-stock portfolio: 24.33% largest decline in plotted observations. 2025-02-13, 16:00:00 EST ($10,415.39) to 2025-04-07, 09:30:00 EDT ($7,881.71).
Show the table · 3 rows, 4 columns
Scroll sideways to compare all columns. Select a heading to sort.
| Published five-stock portfolio | +41.04% | 24.33% | 24.33% |
| New three-stock finalist | +27.66% | 26.53% | 26.53% |
| SPY with matching fees and reinvestment | +30.96% | 19.92% | 19.92% |
No matching rows. Clear the filter to see all records.
Fresh $10,000 starts, 2025-01-01 through 2026-09-30. New comparisons all reinvest dividends and use 0.1% per stock fill with the same 1.5% execution reserve. Both drawdown measurements remain visible.
The published five-stock dividend portfolio returned 41.04% from January 2025 through September 2026, versus 30.96% for SPY with matching fees, reserve and dividend reinvestment. Its drawdown was 24.33%, versus 19.92% for SPY. It grew more and fell further.
The new three-stock finalist returned 27.66% over those same dates, with 26.53% drawdown. Concentration produced much stronger growth in the earlier research period, but that improvement did not carry into the latest comparison.
These are historical simulations with $10,000 starting capital. The original five-stock case study was published on October 10, 2026. This research starts from its actual rules.
Recent risk and exact comparisons
Value (%)
- Published five-stock portfolio 2026-09-30, 16:00:00 EDT
- -4.16%
- New three-stock finalist 2026-09-30, 16:00:00 EDT
- -1.38%
- SPY with matching fees and reinvestment 2026-09-30, 16:00:00 EDT
- -2.02%
All three books use paid dividends reinvested, fractional shares, 0.1% stock fees and the same execution reserve. The complete rule and benchmark names are shown in the adjoining tables.
Inspect the recent drawdown and return bars
Start from the actual published dividend portfolio
The source portfolio selects the five largest S&P 500 companies with ten consecutive completed years of regular dividend increases. It gives each the same weight and reinvests paid dividends. Its refresh rule waits at least thirty days since the last filled buy; it is not a fixed first-trading-day monthly schedule.
The dividend screen compares the closing regular per-share payment rate in completed New York calendar years. It does not add up each year's distributions. Special dividends are excluded and regular payments are split-adjusted.
Show the table · 7 rows, 3 columns
Scroll sideways to compare all columns. Select a heading to sort.
| Dividend raises | Ten completed consecutive years | Five completed consecutive years |
| Stock selection | Five highest market capitalizations | Three highest market capitalizations |
| Position weights | Equal | Proportional to market capitalization |
| Universe | S&P 500; $500M cap floor; one share class per company | Same |
| Refresh | Start, or at least 30 days since a filled buy | Same |
| Dividends | Reinvest paid cash in the paying stock | Same |
| Execution | $10,000; fractional shares; 1.5% reserve; 0.1% per stock fill | Same |
No matching rows. Clear the filter to see all records.
Missing market capitalization is excluded. These size rankings use the historical observations available to the engine.
An earlier exit did not erase Exxon’s loss
The original Exxon exit realized a $426.38 average-cost loss. The three-stock finalist exited earlier, in December 2017, but realized a larger $682.76 loss because of its larger position. Microsoft qualified in January 2015 rather than January 2020.
The losses that shaped the repairs
Inspect Chevron, Microsoft and Exxon order changes
The repair changed which companies qualified and how much they held. Microsoft entered five years earlier; Exxon exited earlier but its realized dollar loss grew. The detailed order comparison is below.
The original portfolio fell 23.23% from its November 2014 peak to its August 2015 trough. Chevron was bought at $110.96 on January 5, 2015, received small additions as its price declined, and was fully sold at $95.06 on July 6. Its average-cost sale return was -14.04% after modeled fees, excluding dividend cash as a separate return component.
Exxon remained in the source book for years and its January 2020 exit realized a $426.38 average-cost loss. The monthly top-ups and trims were ordinary equal-weight maintenance. The original's $141.77 in stock-order fees, plus $2.68 in reinvestment fees, could not explain its roughly 99 percentage-point growth gap to matched SPY.
| Observed change | What the trade record shows |
|---|---|
| Size weights instead of equal weights | Chevron entry shrank from 18.44 to 14.09 shares. Its 2015 entry and exit dates stayed the same. |
| Five-year eligibility instead of ten years | Microsoft entered in January 2015 at $46.37, five years before the original's first entry. |
| Earlier Chevron exit | The five-year book held Chevron in 2014 and exited in January 2015. It avoided the source's later 2015 holding, while retaining an earlier loss. |
| Three stocks instead of five | Microsoft entered January 2015 and Apple January 2018. Training growth rose further, while drawdown increased. |
| Exxon in the three-stock finalist | The main exit moved to December 2017, with a $682.76 average-cost loss. The earlier exit did not remove the loss, and the larger position increased its dollar size. |
No matching rows. Clear the filter to see all records.
Actual filled-order records support these specific changes. A higher total return alone does not show that all losing trades were avoided.
Earlier growth improved as eligibility and concentration changed
The price-trend and momentum screens in the first round changed the holdings frequently. Their growth fell below the original, with some stock-order fees above $800. Adding more rules did not fix the losing episode or the growth gap.
Size weighting raised training growth from 95.66% to 128.95%. Five-year dividend eligibility raised it to 298.08%, with Microsoft entering in 2015. Three holdings then reached 436.03%. These gains came from different holdings and larger positions in the winners, alongside more concentration.
A ninety-day refresh on the five-stock, five-year version returned 343.95%, with 27.92% drawdown, compared with 436.03% and 30.23% for the three-stock finalist. The growth objective selected the latter; the slower five-stock version remains an inspectable tradeoff.
Earlier-period balance paths
Portfolio value (USD)
- Published five-stock portfolio 2021-12-31, 16:00:00 EST
- $19,565.66
- New three-stock finalist 2021-12-31, 16:00:00 EST
- $53,602.74
- SPY with matching fees and reinvestment 2021-12-31, 16:00:00 EST
- $29,487.34
Published five-stock portfolio: 23.23% largest decline in plotted observations. 2014-11-21, 09:30:00 EST ($10,733.30) to 2015-08-25, 16:00:00 EDT ($8,240.24).
Inspect the complete training comparison
Scroll sideways to compare all columns. Select a heading to sort.
| Published five-stock portfolio | +95.66% | 23.23% | 23.23% |
| New three-stock finalist | +436.03% | 30.23% | 30.23% |
| SPY with matching fees and reinvestment | +194.87% | 33.87% | 33.87% |
No matching rows. Clear the filter to see all records.
Fresh $10,000 starts, 2014-01-01 through 2021-12-31. New comparisons all reinvest dividends and use 0.1% per stock fill with the same 1.5% execution reserve. Both drawdown measurements remain visible.
2014 to 2021: the earlier-period selection: drawdowns
Value (%)
- Published five-stock portfolio 2021-12-31, 16:00:00 EST
- -0.57%
- New three-stock finalist 2021-12-31, 16:00:00 EST
- -1.54%
- SPY with matching fees and reinvestment 2021-12-31, 16:00:00 EST
- -0.62%
Inspect training loss paths
All nineteen repair variants
Inspect the three rounds of tested rules
Scroll sideways to compare all columns. Select a heading to sort.
| Five largest, size weights | +128.95% | 24.71% | $97.50 |
| Twenty largest, equal weight | +103.95% | 29.36% | $146.53 |
| Positive free cash flow | +101.76% | 22.84% | $174.15 |
| Twelve-month momentum among twenty large growers | +91.28% | 38.54% | $814.84 |
| Ten largest, equal weight | +80.38% | 27.67% | $126.00 |
| Six-month momentum among twenty large growers | +51.89% | 36.43% | $861.46 |
| Positive six-month price return | +47.65% | 31.29% | $667.01 |
| Positive twelve-month price return | +41.92% | 26.65% | $515.57 |
| Price above 200-day average | +28.39% | 35.39% | $682.19 |
| Price above 100-day average | +28.08% | 37.47% | $828.72 |
| Five-year raises, five size-weighted names | +298.08% | 28.55% | $126.66 |
| Positive cash flow with five size-weighted names | +133.27% | 24.71% | $135.27 |
| Three largest, size weights, ten-year raises | +113.11% | 25.71% | $139.46 |
| Ten largest, size weights | +111.94% | 27.08% | $76.87 |
| Twenty largest, size weights | +111.27% | 27.99% | $73.24 |
| Five-year raises, three size-weighted names | +436.03% | 30.23% | $137.01 |
| Five-year raises, five names, 90-day refresh | +343.95% | 27.92% | $84.63 |
| Five-year raises, five names, positive cash flow | +312.17% | 27.85% | $168.32 |
| Five-year raises, ten size-weighted names | +218.85% | 29.40% | $113.56 |
No matching rows. Clear the filter to see all records.
Three research rounds on January 2014 to December 2021. Every recorded variant is retained.
The earlier-period winner lagged the source on later dates
Portfolio value (USD)
- Published five-stock portfolio 2024-12-31, 16:00:00 EST
- $16,123.74
- New three-stock finalist 2024-12-31, 16:00:00 EST
- $14,083.36
- SPY with matching fees and reinvestment 2024-12-31, 16:00:00 EST
- $12,754.05
Published five-stock portfolio: 17.64% largest decline in plotted observations. 2022-04-21, 09:30:00 EDT ($10,134.52) to 2022-06-17, 09:30:00 EDT ($8,346.91).
Show the table · 3 rows, 4 columns
Scroll sideways to compare all columns. Select a heading to sort.
| Published five-stock portfolio | +61.24% | 17.64% | 17.67% |
| New three-stock finalist | +40.83% | 29.57% | 29.57% |
| SPY with matching fees and reinvestment | +27.54% | 24.71% | 26.24% |
No matching rows. Clear the filter to see all records.
Fresh $10,000 starts, 2022-01-01 through 2024-12-31. New comparisons all reinvest dividends and use 0.1% per stock fill with the same 1.5% execution reserve. Both drawdown measurements remain visible.
On 2022 to 2024 dates, the original five-stock portfolio still led the three-stock finalist. The latest 2025 to September 2026 comparison at the top of this page also favors the original.
2022 to 2024: the published portfolio still leads: drawdowns
Value (%)
- Published five-stock portfolio 2024-12-31, 16:00:00 EST
- -3.76%
- New three-stock finalist 2024-12-31, 16:00:00 EST
- -3.88%
- SPY with matching fees and reinvestment 2024-12-31, 16:00:00 EST
- -3.54%
Inspect the 2022 to 2024 drawdown
Selecting again each period helped overall, with losing folds
The strategy selected on earlier dates can lose to a control on the next dates.
Show the table · 4 rows, 6 columns
Scroll sideways to compare all columns. Select a heading to sort.
| 2014 to 2017 | 2018 to 2019 | Cash flow + momentum | +11.90% | +10.49% | +23.65% |
| 2014 to 2019 | 2020 to 2021 | Three stocks, five-year raises | +102.85% | +44.62% | +49.90% |
| 2014 to 2021 | 2022 to 2023 | Three stocks, five-year raises | +10.96% | +13.41% | +2.34% |
| 2014 to 2023 | 2024 to 2026 (September) | Three stocks, five-year raises | +69.27% | +100.48% | +65.19% |
No matching rows. Clear the filter to see all records.
Returns are cumulative within each test period. The combined result multiplies the fold growth factors; it does not add the percentages.
We froze eight actual candidates and used the same selection rule in four expanding-training folds: choose the highest net growth on the earlier dates, then test it on the next period. The first fold selected Aurora's cash-flow and momentum variant. The next three selected the three-stock, five-year version.
Compounding the independently restarted test paths produced 326.33% growth from January 2018 through September 2026, versus 263.31% for the original and 213.36% for matched SPY. The procedure lost to SPY in the first fold and to the original in the last two folds.
The candidate family was devised after earlier research had already viewed these historical dates. This is a retrospective walk-forward comparison. Each fold restarts from cash, includes its first-session move and entry fee, and does not charge a hypothetical liquidation at the boundary.
Retrospective selection procedure: chained balances
Portfolio value (USD)
- Retrospective selection procedure 2026-09-30, 16:00:00 EDT
- $42,632.67
- Published five-stock portfolio 2026-09-30, 16:00:00 EDT
- $36,330.57
- SPY with matching fees and reinvestment 2026-09-30, 16:00:00 EDT
- $31,335.94
Retrospective selection procedure: 30.76% largest decline in plotted observations. 2020-02-11, 09:30:00 EST ($12,527.29) to 2020-03-23, 16:00:00 EDT ($8,673.93).
The original and SPY are restarted at the same boundaries. The plotted fold-start points mark the declared starting capital before that fold's trading begins.
Retrospective selection procedure: chained drawdowns
Value (%)
- Retrospective selection procedure 2026-09-30, 16:00:00 EDT
- -1.37%
- Published five-stock portfolio 2026-09-30, 16:00:00 EDT
- -4.16%
- SPY with matching fees and reinvestment 2026-09-30, 16:00:00 EDT
- -2.03%
Inspect chained drawdown and exact totals
Scroll sideways to compare all columns. Select a heading to sort.
| Retrospective selection procedure | +326.33% | 30.76% |
| Published five-stock portfolio | +263.31% | 24.33% |
| SPY with matching fees and reinvestment | +213.36% | 33.65% |
No matching rows. Clear the filter to see all records.
Historical paths from the same four test periods, January 2018 through September 2026.
The earlier yield search failed
Portfolio value (USD)
- Earlier twenty-stock yield screen 2026-09-30, 16:00:00 EDT
- $10,983.67
- Earlier SPY: 95% initial stocks, cash dividends 2026-09-30, 16:00:00 EDT
- $12,964.60
- Earlier SPY: 100% initial stocks, reinvest, no fill fees 2026-09-30, 16:00:00 EDT
- $13,153.73
Earlier twenty-stock yield screen: 16.20% largest decline in plotted observations. 2025-03-03, 16:00:00 EST ($10,471.54) to 2025-04-09, 09:30:00 EDT ($8,775.26).
Inspect the earlier yield-search result
Scroll sideways to compare all columns. Select a heading to sort.
| Selected twenty-stock, six-year yield screen | +9.84% | 16.20% |
| Earlier three-year yield screen | +18.62% | 18.19% |
| SPY: 100% initial stocks, reinvest, no fill fees | +31.54% | 20.19% |
| SPY: 95% initial stocks, cash dividends, 0.05% fee | +29.65% | 19.22% |
| SCHD: 95% initial stocks, cash dividends, 0.05% fee | +22.80% | 14.70% |
No matching rows. Clear the filter to see all records.
The earlier controls have different cash and dividend conventions from the new source-based experiment. Compare within each experiment.
The earlier thirty-configuration search selected twenty stocks by highest dividend yield, with six completed years of dividend increases and indicated yields between 1% and 6%. That book returned 9.84% in January 2025 through September 2026. It used cash dividends, a 95% stock-deployment target subject to the execution reserve, and 0.05% stock fees.
Its fully invested, dividend-reinvesting SPY companion returned 31.54% and excluded modeled fill fees. The explicit SPY control started with 95% stocks, charged the same 0.05% entry fee and left dividends as cash; it returned 29.65%. The earlier result remains visible rather than being replaced by the more favorable five-stock source.
Aurora's separate tested alternatives
Read the native Aurora comparisons
Scroll sideways to compare all columns. Select a heading to sort.
| Ninety-day refresh | +80.89% | 26.56% |
| 95% deployment | +89.58% | 22.16% |
| Sixty-day refresh | +98.40% | 23.22% |
| 98.5% deployment | +93.82% | 22.91% |
| Sixty-day refresh and 95% deployment | +92.16% | 22.15% |
| Trend and momentum, ten stocks | +55.12% | 34.89% |
| Cash flow and momentum among twenty large growers | +106.68% | 37.26% |
| Final trend guard | +212.07% | 30.18% |
| Three-year eligibility, five size-weighted stocks | +367.19% | 28.55% |
No matching rows. Clear the filter to see all records.
All nine actual alternatives use the source cost and dividend convention. Unchanged clones and failed authoring requests are excluded.
Aurora inspected the trades; the engine calculated the results
Read the agent and engine workflow
The actual native Aurora session inspected the original Chevron and Exxon order records and proposed refresh, reserve, trend, cash-flow and momentum changes. Codex saved the additional declared variants through NexusTrade MCP, compared them, retrieved actual fills, and ran the fixed-family folds. NexusTrade's engine produced the backtests.
Aurora's combined trend and momentum repair returned 55.12% in 2014 to 2021; its cash-flow and momentum repair returned 106.68%. Both trailed matched SPY in the full training period. The latter won the shorter first-fold training period and then returned only 11.90% against SPY's 23.65% in 2018 to 2019.
Two authoring requests failed to produce the requested rules: a churn guard and a position-weight cap. Those failures and unchanged duplicate drafts were retained rather than counted as new tested ideas. All portfolios remain inactive.
How to read these comparisons
Read dates, fees, reinvestment and measurement details
New results compare actual holding books with the same $10,000 start, fractional shares, 1.5% execution reserve, 0.1% stock fees and dividend reinvestment. Paid dividends buy additional shares of the paying stock. There is no added spread, market-impact or tax model.
Returns use ending NAV divided by the original $10,000, including the first trading session and fees. Drawdown uses all exported native NAV observations, including open, close and corporate-action timestamps. Charts show recorded bucket extrema and exact endpoints; complete histories remain in the source export. The charts do not measure every intraday extreme.
The original published article ended on October 6, 2026 and reported 44.80% for its recent start. This page reran common dates ending September 30 and reports 41.04%. The different endpoints explain why those recent return numbers are not interchangeable.
A future period observed after these choices, or a new paper test, would add prospective evidence. The latest historical comparison already favors the original five-stock book.
No comments yet.