docs: freeze Regime Trend v1 profit protection matrix

Author: dharmanan

research/regime-trend-v1/profit-protection-plan-v1.md

@@ -0,0 +1,122 @@
+# Regime Trend v1 — Frozen Profit Protection Matrix
+
+Status: **Frozen before counterfactual overlay results are viewed**
+
+Purpose: test whether completed-bar profit protection can improve the frozen 4-hour strategy without destroying its large right-tail winners.
+
+This stage is counterfactual research. It does not change the baseline strategy and does not open the final holdout.
+
+## 1. Data boundary
+
+Evaluate the existing partitions separately:
+
+- Development: 2019-01-01 through 2022-12-31 UTC
+- Validation/research: 2023-01-01 through 2024-12-31 UTC
+
+The final holdout beginning 2025-01-01 remains closed.
+
+## 2. Activation rule
+
+A protection rule activates only after a completed 4-hour candle closes at or above:
+
+`entry_fill + activation_atr * frozen_entry_atr`
+
+The protection floor becomes active on the next 4-hour candle. It never acts inside the activation candle.
+
+The actual exit candle close is not eligible to activate protection because the baseline position may already have exited before that close.
+
+## 3. Net floor definition
+
+Each floor is specified as a desired net result in frozen entry-ATR units after normal exit slippage and commission when the floor is touched exactly.
+
+The raw stop reference is solved from:
+
+- baseline entry fill,
+- baseline quantity,
+- already-paid entry commission,
+- exit commission of 0.10%,
+- exit slippage of 0.05%.
+
+A gap below the floor exits at the actual candle open with adverse slippage; therefore realized net result may be below the desired floor.
+
+## 4. Frozen candidates
+
+Only these seven candidates are allowed:
+
+1. `activate-0.50-lock-0.00`
+   - activation: +0.50 ATR completed-bar close
+   - net floor: 0.00 ATR
+
+2. `activate-1.00-lock-0.00`
+   - activation: +1.00 ATR completed-bar close
+   - net floor: 0.00 ATR
+
+3. `activate-1.00-lock-0.25`
+   - activation: +1.00 ATR completed-bar close
+   - net floor: +0.25 ATR
+
+4. `activate-1.00-lock-0.50`
+   - activation: +1.00 ATR completed-bar close
+   - net floor: +0.50 ATR
+
+5. `activate-1.50-lock-0.00`
+   - activation: +1.50 ATR completed-bar close
+   - net floor: 0.00 ATR
+
+6. `activate-1.50-lock-0.50`
+   - activation: +1.50 ATR completed-bar close
+   - net floor: +0.50 ATR
+
+7. `activate-1.50-lock-1.00`
+   - activation: +1.50 ATR completed-bar close
+   - net floor: +1.00 ATR
+
+No additional trigger or floor may be introduced after viewing results.
+
+## 5. Conservative execution policy
+
+For candles after activation and before the baseline exit candle:
+
+- if open is at or below the floor, exit at open with adverse slippage,
+- otherwise if low reaches the floor, exit at the floor with adverse slippage.
+
+On the baseline exit candle, only an opening gap through the floor may create an overlay exit. The later high/low path of that candle is not used because baseline exit ordering is already in force and intrabar sequence is unknown.
+
+If the overlay does not produce a provably earlier exit, the original baseline trade is retained unchanged.
+
+## 6. Required outputs
+
+For every candidate and partition report:
+
+- baseline and overlay total net PnL,
+- change in total net PnL,
+- baseline and overlay profit factor,
+- baseline and overlay win rate,
+- baseline and overlay maximum drawdown,
+- activation count,
+- earlier overlay-exit count,
+- losing trades converted to winners,
+- losing trades improved but still losing,
+- winners reduced,
+- winners converted to losses,
+- unchanged trades,
+- median, 90th-percentile and maximum realized net ATR,
+- number of baseline winners preserving at least 90% of their original net PnL,
+- results by symbol.
+
+## 7. Interpretation discipline
+
+A candidate is not accepted because it has the highest validation result.
+
+A viable profit-protection hypothesis must:
+
+- improve development and validation/research total net PnL,
+- not reduce profit factor in either partition,
+- not increase maximum drawdown in either partition,
+- preserve at least 80% of baseline winning trades at 90% or more of their original net PnL,
+- avoid converting more than 5% of baseline winners into losses,
+- show improvement across more than one symbol.
+
+If no frozen candidate satisfies these conditions, completed-bar fixed-floor protection is rejected and the next research step becomes a separately specified lower-timeframe exit overlay or entry/no-trade filter.
+
+No final holdout access is allowed during this stage.