Name the instrument and the data file. Times are New York time, prices are rounded to the tick, and contract roll weeks are removed. The same cleaned data is used in every window.
| Item | Value |
|---|---|
| Instrument | NQ front month, 1 contract, $20 per point |
| Data file | /workspace/setuphunter_py/data/nq_front_month_m1.csv (1-minute bars, 407,551 bars after cleaning) |
| Time zone | Converted from UTC bar end to New York time bar start. Sessions are set in UK clock time and converted bar by bar. |
| Tick grid | All prices rounded to 0.25. Check on 81 Full Mech trades (324 entry/stop/target/exit prices): 0 off the 0.25 grid. |
| Roll weeks removed | 8–14 Sep 2025; 10–12 Dec 2025; 11–13 and 15 Mar 2026; 11–12 Jun 2026 |
| Other cleaning | Sunday pre-open bars dropped |
| Same data every window | Yes: one load (flib.load()) feeds the tuning, out-of-sample and holdout runs |
No roll-week or tick-grid charts. Roll weeks are removed before the run, so no trade or day chart falls inside one, and no price was off the tick grid, so there is no failure to show. Data gap that did matter: the file starts 30 Jun 2025 22:41 ET, which cut off the Asia low behind Berg's verified entry GT01 (see rule 8).