Read the full methods, uncertainty and results report
Explore how the same rules change with implementation costs, taxes, capital and a deliberately biased universe. Every evaluated rule is shown, including losses. Empty windows and failures are published in the evidence files.
| Rule | Dates | Months | Status | CAGR | Max drawdown | Annualized Sharpe (RF=0) | Capacity breaches |
|---|
These records belong to the selected implementation and data treatment. A missing result must not be read as a zero return.
Gross uses full monthly rebalancing; implemented portfolios use a 25% relative rebalance band, so their difference includes implementation-policy effects. Tax results use a simplified resident-individual FIFO model. A 10% single-day ADV breach marks a capacity concern; the model does not simulate partial fills or trading circuits. Costs at larger capital do not prove executable capacity.
Deleting stocks that disappear uses future knowledge. This is an educational bias experiment, never an investable strategy. A positive return or a Sharpe ratio is not proof of alpha. The historical 2018 split has already been inspected; it is not a fresh holdout.
“Assume last mark” cashes out unavailable observations at the carried mark; “assume total loss” writes off existing holdings. Both reject new target fills when their future holding return is missing. These use hindsight as audit sensitivities and do not establish a causal, executable exit policy. Strict implementations publish missing-price failures.
Value means point-in-time fundamentals. Long-term reversal and price relative to a five-year mean are price signals, not substitutes for book-to-market or earnings yield. Full-history fundamental value remains untested here.
No trades are sent. No broker credentials are used. This is a research and education tool.