Frozen protocol / version 1.0
Research dossier 001
Synthetic securities and returns. A reproducibility exercise, with no claim of a historical equity premium.
Synthetic sensitivity analysis. Changing settings after viewing the fixture does not create a new independent holdout.
Period ledger
| Period | Split | Selected | One-way turnover | Stock cost | Gross | Net |
|---|
Cost sensitivity
Protocol and information clock
The universe contains six fictional securities and an explicit cash balance. Scores represent a supplied research input; they are not fitted from the example returns. At each formation cut-off, every eligible score must already be available. Ties are resolved deterministically by security identifier.
January is labelled development, February validation and March holdout within the synthetic clock. The dataset was authored on 16 October 2026. Its earlier dates are illustrative and do not claim that a model or forecast existed in 2023 or 2024. These tiny synthetic samples support arithmetic verification only.
Weights are long-only. The maximum security weight can leave cash uninvested. Stock buy and sell notionals incur the supplied basis-point rate; the cash leg is shown in one-way turnover but is not separately charged as a stock trade. Rebalancing starts from the drifted holdings. Read the fixture README for the exact simplified fee convention.
The fixture supplies both price and total-return fields. The dossier uses total return. Corporate actions, distributions, index membership and actual publication times would require dated, licensed evidence in a real study. No historical trading result can be inferred from this fixture.