Independent audit · October 12, 2026
Put $1,000 into the strongest S&P 500 stock each month. Keep every purchase. Compare it with putting the same money into a broad market fund.
January 2, 1998 – October 9, 2026 · 346 monthly deposits · US dollars
Explore the recorded portfolio value over time. Each line includes the money added along the way.
Changing the deposit scales the same historical purchases; it does not rerank stocks. The final point is October 9, not the end of October.
Apple and Nvidia account for 71.4% of the ending portfolio. Eight deposits into those two companies grew to about $5.86 million in this model.
Most individual purchases did not beat the same $1,000 invested in SPY. A small number of very large gains drove the total.
The 113 zero-value purchases are included in the 155 losses. Missing histories, acquisitions, and bankruptcies all receive $0 in the supplied model.
To know which stock won, you need every eligible stock’s return. In the first month, the model could rank only 264 of 499 members.
Choosing among the available names can favor companies with surviving price histories. Writing off a bought company later does not restore a missing candidate at the time of selection.
Each square is a purchase month. Select one to inspect its prior-month ranking coverage.
An independent rebuild reproduces the supplied approximation. It does not validate the unobserved stocks or prove the result could have been achieved in a real account.
All 346 purchases, observed rankings, and monthly portfolio values were recomputed from the frozen inputs. Totals match to within a cent: $8,204,466.91 and $2,258,566.73. All four supplied input checksums match. Fresh Yahoo checks also agree with the SPY, Apple, and Nvidia purchase prices; this is a cross-check against a shared upstream provider, not a second independent market feed.
Only 111 rankings have complete price coverage within the supplied membership list. The $8.20M result belongs to the best-observed-stock approximation. A complete historical stock universe and corporate-action history are needed to test the original claim.
The code values any purchased series ending before October 9, 2026 at zero. That includes acquisitions and inaccessible histories. It writes off 113 purchases in 72 stocks, originally costing $113,000. The historical line drops a holding only after its final quote.
The model uses members at the previous month’s end. JCP was selected for December 2, 2013, its removal date in the supplied file. Four split corrections are documented heuristics; the archive lacks the daily corporate-action evidence needed to independently certify each one. Dividend and spin-off returns follow the provider’s adjustments.
17.84% and 11.11% are money-weighted annual returns, accounting for the dates money went in. They use a 365.25-day year. Standard Excel XIRR uses 365 days and gives 17.83% and 11.11%. The workbook’s return cells are fixed numeric formulas, so they will not recalculate if the underlying results change.
Rank the preceding calendar month’s total return among members at signal month-end. Buy the observed leader at the next month’s first trading-day open; ties use ascending ticker order. Repeat winners receive another $1,000. Hold purchases even after they leave the S&P 500. The comparison buys SPY, an exchange-traded fund tracking the S&P 500, at those same opens.
Both sides use fractional shares and reinvested distributions, with no taxes, commissions, spread, or slippage. SPY fund costs are embedded. The adjusted open is open × adjusted close ÷ close; adjusted units are bookkeeping units, not literal shares. No selection signal comes after its buy date. That chronological check does not prove the reconstructed data were available in this exact form at the time.
| Scenario | Money added | Strategy | SPY |
|---|---|---|---|
| Original approximation | $346,000 | $8,204,467 | $2,258,567 |
| Also lose every deposit in an incomplete-ranking month | $346,000 | $419,574 | $2,258,567 |
| Buy only in the 111 complete-ranking months | $111,000 | $419,574 | $258,582 |
The second row is deliberately harsh, not a lower bound on the true strategy. The third uses a different contribution schedule on both sides. Neither repairs missing history or gives a confidence interval.
What each $1,000 deposit became by October 9, 2026.
| Bought | Stock | Prior month | Ending value | Same deposit in SPY |
|---|
This table always uses the original $1,000 deposits. A date suffix in a stock identifier distinguishes a historical series.
Download the supplied data, inspect the audit, or reproduce the approximation. The Python Agents API runner uses gpt-6.1-sol with medium reasoning and downloads the files from the completed run.
Includes the original workbook and source archive, frozen price inputs, all rankings, a clear research prompt, Python scripts, and .env.example. The delivered audit was reproduced locally; the hosted API workflow was tested without making a paid live run.