Portfolio Forward Replay
Shared prop-firm router test for the Open Drive Quality anchor plus four promoted add-on sleeves. This is a conservative prop-survival anchor, not a scalper: signals are accepted, skipped, canceled, or bracketed under one global risk governor.
Bottom line
Best pure payout-survival combo: Funded Safe. It is the cleanest governor for staying inside modeled funded-account constraints, but it is too slow to be the main daily-trading strategy.
| Router mix | Eval pass | First funded payout | Pass + payout read | Median eval | Trade flow |
|---|---|---|---|---|---|
| Funded Safe | 99.7% | 100.0% | 99.7% | 80 trading days | 1.04 trades/week |
| Research All Signals | 100.0% | 99.9% | 99.9% | 47 trading days | Diagnostic only |
| Evaluation Balanced Two-Trade | 99.9% | 55.1% | 55.0% | 62 trading days | 1.31 trades/week |
| Evaluation Strict | 100.0% | 0.0% | 0.0% | 57 trading days | 1.04 trades/week |
My practical recommendation: use Funded Safe as the payout-survival benchmark, not the first strategy to automate for a quick evaluation pass.
Use Evaluation Balanced only in demo if the goal is to study eval speed. It gives up too much funded payout quality to be the default funded profile.
Do not deploy Research All Signals as a live profile. It is useful for research because it shows available edge, but it does not enforce the same practical trade cap.
One portfolio, different governors.
Research All is diagnostic and fastest because it has no daily trade cap. The practical split is Evaluation Balanced for faster eval attempts and Funded Safe for payout survival.
| Mode | Accepted | Win | ExpR | PF | Max DD | Eval pass | Eval fail | Eval median | Funded payout |
|---|---|---|---|---|---|---|---|---|---|
| Funded Safe | 161 | 68.3% | +0.411R | 2.44 | 4.44R | 99.7% | 0.0% | 80.0 | 100.0% |
| Research All Signals | 247 | 64.4% | +0.321R | 1.96 | 6.39R | 100.0% | 0.0% | 47.0 | 99.9% |
| Evaluation Balanced Two-Trade | 203 | 65.5% | +0.345R | 2.08 | 4.52R | 99.9% | 0.0% | 62 | 55.1% |
| Evaluation Strict | 161 | 68.3% | +0.411R | 2.44 | 4.44R | 100.0% | 0.0% | 57 | 0.0% |
Use this as the conservative anchor, not the daily trader.
This profile is intentionally selective: roughly 1.04 trades per week in the funded-safe mode. It is useful for prop-survival research, but the next alpha hunt should target a separate higher-frequency sleeve instead of loosening this governor until it becomes something it was not tested to be.
| Profile | Best use | Key numbers | Constraint |
|---|---|---|---|
| Evaluation Balanced Two-Trade | Demo eval-speed profile | 203 trades, 65.5% win, 99.9% eval pass, median 62d | Allows second trade only if no open position and no daily-loss lock. |
| Funded Safe | Funded/payout-survival profile | 161 trades, 68.3% win, 1.04 trades/week, 100.0% payout read, median 21.0d | Uses $110 risk so one winning day can clear the modeled payout-day threshold. Too slow for a standalone scalper. |
| Evaluation Strict | Lowest-complexity baseline | 161 trades, 68.3% win, 100.0% eval pass | One trade/day; slower but simpler to audit. |
| Research All Signals | Diagnostic only | 247 trades, 64.4% win, 100.0% eval pass | No daily trade cap, so do not treat as the first deploy profile. |
Missed entries are canceled, not chased.
The synthetic missed-fill test canceled every seventh submitted strict-mode order. It still accepted 154 trades, canceled 25 stale orders, and modeled 100.0% eval pass with median 60d.
Execution state machine now has a replay-matching harness.
The harness converts the selected signal ledger into adapter-style order states. Both practical profiles matched the portfolio replay exactly and passed all invariant checks: brackets before close, no stale fills, no overlapping positions, and daily caps respected.
| Harness run | Submitted | Accepted | Skipped | Stale cancels | QA | Replay match |
|---|---|---|---|---|---|---|
| Evaluation Balanced Two-Trade | 203 | 203 | 177 | 0 | pass | exact |
| Funded Safe | 161 | 161 | 219 | 0 | pass | exact |
| Balanced stale test | 213 | 183 | 167 | 30 | pass | synthetic misses |
| Funded stale test | 179 | 154 | 201 | 25 | pass | synthetic misses |
Local replay files behind this page
This page is based on local strategy tradebooks and ledgers only. No paid data was pulled for this update.
| Area | Source | Use |
|---|---|---|
| Candidate set | open_drive_quality, MPS_6bd769ceff, MPS_72f259c691, MPS_05cbcc6310, MPS_d93bd83557 | Open Drive Quality plus four sleeves promoted by deterministic and targeted prop tests. |
| Router report | work/prop_alpha_portfolio_router_replay_20260628/PORTFOLIO_ROUTER_REPLAY.md | Mode table, files, and why this gate came before another broad search. |
| Event logs | work/prop_alpha_portfolio_router_replay_20260628/portfolio_router_events_*.csv | Signal confirmed, submitted, filled, bracket attached, skipped, stale canceled, and exited states. |
| Trade logs | work/prop_alpha_portfolio_router_replay_20260628/portfolio_router_trades_*.csv | Accepted routed trades for each governor mode. |
| Paper harness | work/prop_alpha_paper_router_harness_20260628/PAPER_ROUTER_HARNESS_REPORT.md | Adapter-facing state machine QA with exact replay match for practical profiles. |
| Stale harness | work/prop_alpha_paper_router_harness_stale_20260628/PAPER_ROUTER_HARNESS_REPORT.md | Synthetic missed-fill test proving stale orders cancel instead of chasing. |
| Prop rules | Topstep-style 50K template last verified 2026-06-25 | Must be refreshed from official firm rules before any real challenge. |
Risk note
Strategy research is educational software and trading research, not financial advice. This is a paper/demo forward-replay candidate, not a live-funded approval. Live markets can break backtests through fees, slippage, latency, platform behavior, missed fills, rule changes, and regime shifts.