The exclusion uses candidate id plus pre-entry minute bucket. It does not use MAE, MFE, final result, final R, or hold-time diagnostics as entry filters.
Current best demo candidate
MNQ/MES DBP Morning Filter Router
A 556-trade hierarchical opening-structure router. It keeps the current detector-ready anchor but blocks one DBP open-drive reclaim component during 08:00-08:59 entry time, because that slice increased prop-firm drawdown tail risk.
Trades5562023-06-01 to 2026-05-27
Win rate60.8%338 wins / 218 losses
Net return+213.0Rclosed-trade R units
Expectancy+0.383Rper closed trade
Profit factor2.06gross wins / gross losses
Max drawdown5.63Rhistorical path
Frequency3.56/wk1.62 per active day
SymbolsMES 56, MNQ 500tradebook split
Bottom line
Best current profile: use the safe/default governor for funded-style survival and paper observation. Use the faster governor only as an evaluation-mode A/B profile after live paper evidence exists.
| Profile | Risk | Daily locks | Max trades | 44d pass/fail | 66d pass | Median/p75 | Funded60 payout | Decision |
|---|---|---|---|---|---|---|---|---|
| Safe/default | $400 | $800 profit / $350 loss | 4/day | 74.33% / 6.29% | 88.09% | 26d / 38d | 98.69% | Preferred paper profile |
| Faster eval A/B | $450 | $1000 profit / $500 loss | 2/day | 78.88% / 8.65% | 88.21% | 23d / 34d | 98.57% | Faster but rougher stress tail |
Equity Curve Preview
556 filtered trades
Reality gate
What still has to happen
- Load current data access and rerun the launch gate.
- Observe one fresh accepted signal with source entry minute visible.
- Send exactly one supervised paper bracket before any paid evaluation order.
- Keep private exact code and thresholds locked until checkout/access flow exists.
Audit trail
Why this was promoted
The live detector now exposes source signal minute and source entry minute, and DBP9 is blocked only for 08:00-08:59 source entry time.
The safe governor improved failure tail versus the prior 608-row anchor while preserving funded-mode payout survival.
Site candidate and demo candidate. Not automation candidate for paid Topstep/eval until live paper proof exists.
Risk note
Strategy research is educational software and trading research, not financial advice. Backtests and Monte Carlo simulations can fail in live markets because of fees, slippage, latency, platform behavior, missed fills, rule changes, and regime shifts.