docs: record Regime Trend v1 execution errata

Author: dharmanan

research/regime-trend-v1/errata-001.md

@@ -0,0 +1,41 @@
+# Regime Trend v1 — Errata 001
+
+Status: **Accepted before any market-data evaluation**
+
+This errata corrects two contradictions discovered by synthetic-test design. No market data or performance result was viewed before these corrections.
+
+## 1. Remove unreachable immediate-entry-stop rule
+
+The frozen specification stated that an entry candle could open below the newly computed initial stop.
+
+That state is unreachable because:
+
+- `entry_fill = open * (1 + buy_slippage)`
+- `initial_stop = entry_fill - 2.5 * entry_atr`
+- `entry_atr > 0`
+
+Therefore `initial_stop` is always strictly below `entry_fill`, while `entry_fill` is above the raw open. The raw open cannot be less than or equal to the new initial stop.
+
+Corrections:
+
+- remove `immediate_entry_stop` from exit reasons,
+- remove `immediate_entry_stop_flag` from the ledger,
+- the entry candle is handled by the normal initial-stop rule after the entry fills.
+
+## 2. Correct next-open trend-exit ordering
+
+A trend-exit signal is created at candle `t` close and must execute at candle `t+1` open.
+
+Correct ordering at `t+1` open:
+
+1. If `open[t+1] <= active_stop`, classify the fill as a gap stop and use adverse sell slippage from the open.
+2. Otherwise fill the pending trend exit at the same open with adverse sell slippage.
+3. Do not inspect the later intrabar low before executing an already pending next-open trend exit.
+
+The earlier reference engine incorrectly checked the full low of `t+1` before the pending trend exit, which could reclassify a required open exit as an intrabar stop.
+
+## Version impact
+
+- Strategy ID remains `regime-trend-v1` because these are logical corrections made before market evaluation.
+- Reference implementation version becomes `typescript-reference-v1.0.1`.
+- All later implementations and parity reports must apply this errata together with `spec.md`.