Follow one original share
The invented SYN-01 starts its first interval at 100. A two-for-one split turns one original share into two shares, each ending at 54. End share value per original share is therefore 108. The interval also includes a distribution of 2 per original share, paid at the endpoint. The raw change 54/100-1 = -46% compares inconsistent share units and is not the investor's loss.
Split-adjusted price return is 108/100-1 = 8%; total return is (108+2)/100-1 = 10%. The distribution is treated as reinvested at the endpoint, before any later period begins. There are no taxes or intra-period reinvestment assumptions. S&P DJI's methodology distinguishes total returns and corporate-action adjustments; these invented numbers are not index constituent data or a replication of that methodology.
Keep the portfolio comparison consistent
The default first-period selected price returns are 8%, 4% and -3%; total returns are 10%, 4% and -2%. Equal thirds therefore yield a 3% price-only gross return and a 4% total gross return. Selection is unchanged. The price-only view intentionally omits distributions and cannot be described as total investor wealth.
The stored corporate-action values must reconcile to both stored return fields. Replacing the split ratio with one while leaving the returns unchanged raises INVALID_CORPORATE_ACTION. The engine will not repair a mismatch by selecting whichever return makes the result look stronger. General real-world actions such as mergers, rights, spin-offs and complex delistings are outside this small fixture.
The universe is part of the experiment
The six synthetic members are fixed before any formation, including negative-return observations. Removing SYN-03 because it lost money in the first period fails the six-row requirement with MISSING_DATA. The engine does not quietly fill the vacancy with the next-ranked security. An absent return also fails even when the affected name is not selected.
A real survivor-only database can change the eligible sample, but this toy fixture cannot estimate that bias: it has no genuine listings or delistings. Its narrow lesson is reproducibility. A changed universe, corporate-action convention or missing-data policy is a changed experiment and must be stated, not hidden behind the same headline return.
Inspect the example
| Measurement or input | Calculation / condition | Result |
|---|---|---|
| SYN-01 raw price change | 54/100 - 1 | -46%; inconsistent share units |
| SYN-01 split-adjusted price return | 2*54/100 - 1 | 8% |
| SYN-01 total return | (2*54+2)/100 - 1 | 10% |
| Selected portfolio price / total gross | Equal thirds, same three selections | 3% / 4% |
| Remove SYN-03 from first period | Five observations for six fixed members | MISSING_DATA; no result |
Sources & reproduction
Primary references support the definitions and method. Numerical examples are illustrative unless a source record is explicitly identified.