Historical rule test
Refine research. Do not replace it. This is hygiene for a rule the user
already stated, not a strategy generator.
- Write the rule, universe, window, and rebalance before any return is calculated.
- Use only information visible at each decision date. Forward returns must start after that date. Do not fill a statement or estimate into an earlier bar.
- Compute the test with
compute_runorfinance_calculate, supplying the current-turn series explicitly. Any successfully delivered read result is usable; identify its provider, warnings, and coverage. - Disclose sample size, missing bars, costs or frictions if they would change the result, and any look-ahead you could not remove.
- Do not present the result as a live signal, rating, or order.
Lead with whether the stated rule held in the tested window, then the limits.