ashare-sop-l4-execution-review

name	ashare-sop-l4-execution-review
description	L4 execution & review engine for the A-share SOP. Use to convert L3 candidate signals into trade plans (with stop/target/R:R), perform manual confirmation, journal trades, and run weekly review & attribution.
version	1.0.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["a-share","sop","execution","review","l4","trade-plan","journaling"],"related_skills":["a-share-theme-structure-sop","ashare-sop-l3-stock-resonance"]}}

A-Share SOP L4 — Execution & Review Engine

Overview

This is the L4 tier of the A-share theme structure SOP engine (v1.7+). It converts L3 candidate signals into actionable trade plans, enforces manual confirmation (人工终裁), and journals all outcomes for iterative engine improvement.

This tier is human-executed by design. The engine provides the framework and templates; the trader provides the final decision.

Load this skill when:

• L3 has output "候选观察" and you need to build trade plans
• Conducting post-trade review or weekly复盘
• Setting up trade journaling infrastructure

Engine Position

L1 → L2 → L3 (候选观察/风控观察) → L4 (本模块: 交易计划/确认/复盘)

L4.1 Trade Plan Generation (盘后)

When L3 outputs "候选观察", complete these steps before entering manual confirmation:

Step 1: 基本面定性检查 (Fundamental Quick Check)

Reference the BMP value investing framework. Quickly assess:

• 行业前景: Does the industry have clear growth prospects?
• 业务模式: Is the business model understandable and defensible?
• 硬伤排除: Obvious red flags (regulatory risk, accounting concerns, structural decline)

Output: Pass / Fail. If Fail, remove from candidate list with reason noted.

Step 2: 交易计划制定 (Trade Plan Construction)

For each candidate that passes the fundamental check:

Element	Description	Standard
剧本类型 (Setup Type)	趋势转折 / 回踩趋势 / 密集区突破	Must match the L3 signal context
入场触发条件 (Entry Trigger)	Specific intraday price signal, more precise than L3 alert	e.g. "突破 XX.XX 元且 15min K线收阳"
防守位/止损位 (Stop Loss)	Structural low, key MA, or volume node	Must be clear and non-ambiguous
第一目标位 (First Target)	Prior high, platform resistance, or overhead supply zone	Realistic, data-backed
计划盈亏比 (Planned R:R)	(Target - Entry) / (Entry - Stop)	MUST exceed 3:1
初始计划亏损金额 (Planned Max Loss)	Position size × (Entry - Stop)	Calculated from total account; must be within risk tolerance

Entry Trigger Examples

Setup	Trigger Example
趋势转折	放量突破20日高点，15min K线收阳确认
回踩趋势	回踩MA20不破，出现下影线+放量反弹
密集区突破	突破平台整理上沿，伴随量能>5日均量1.5倍

L4.2 Manual Confirmation & Execution

所有交易计划必须经过人工确认。严禁自动化交易。

Confirmation Process (确认流程)

Record in the trade journal:

1. 确认时间 (Confirmation Time): YYYY-MM-DD HH:MM
2. 确认前提 (Confirmation Premise): What signals were checked (L1 state, L2 stage, L3 scores, fundamental check)
3. 确认结论 (Confirmation Conclusion): Execute / Adjust / Abandon, with reason

Manual Execution Boundaries

• Permissible deviation from planned entry: ±0.5% of planned price
• If entry price deviates beyond boundary → re-evaluate R:R before executing
• If stop/target levels are invalidated by gap opening or news → abort and re-plan

L4.3 Trade Journaling & Review

Sample Entry (样本录入)

All completed trades (win/loss) AND important untriggered signals must be recorded with:

• 交易前状态 (Pre-Trade State): L1/L2/L3 readings, plan details, entry trigger status
• 计划 (Plan): Full trade plan from L4.1
• 执行过程 (Execution Process): Actual entry/exit prices, any deviations
• 结果 (Outcome): P&L, R-multiple achieved
• 归因 (Attribution):
  • ✅ Rule-adherent win: followed plan, won
  • ✅ Rule-adherent loss: followed plan, lost (acceptable)
  • ❌ System-external error: emotional override, FOMO, revenge trading
  • ⚠️ Rule gap: situation not covered by existing rules

Weekly Review (每周复盘)

Using the trend trading system weekly review template:

Metric	Calculation
胜率 (Win Rate)	Winning trades / Total trades
盈亏比 (Profit Factor)	Total gains / Total losses
平均R倍数 (Avg R-Multiple)	Average (outcome / planned risk)
规则内失败 (Rule-Adherent Losses)	Count + total loss
系统外错误 (System-External Errors)	Count + identify pattern
规则缺口 (Rule Gaps)	Situations not covered → propose rule update

Review Output

## 本周复盘 YYYY-MM-DD 至 YYYY-MM-DD

### 交易统计
- 总交易：X 笔
- 胜率：XX%
- 盈亏比：X.XX
- 平均R倍数：X.XX

### 归因分析
- ✅ 规则内盈利：X 笔 (+X.XX R)
- ✅ 规则内亏损：X 笔 (-X.XX R)
- ❌ 系统外错误：X 笔 (-X.XX R)
- ⚠️ 规则缺口：X 笔

### 改进措施
1. ...
2. ...

### 下周易错点提醒
- ...

Supporting Infrastructure

The L4 process is manual by design, but supported by automated tracking:

Tool	Function	Usage
performance_tracker.py	Auto-tracks D1/D3/D5 forward returns of L3 candidates	Daily cron f7220d53fa23 at 20:30 Beijing
tracker_index.csv	Cumulative record of candidate outcomes	Read for weekly review stats
historical_sop_replay.py	Backtests candidate rules against historical data	Validate rule changes before adopting
Obsidian vault templates	Trade journal and weekly review templates	Manual entry

Operating Rules

1. 盘后执行 (Post-Market Only): Trade plan generation happens after market close. No real-time plan adjustment during trading hours.
2. 计划盈亏比硬性门槛: R:R < 3:1 → auto-reject, do not pass to confirmation.
3. 每周复盘不可跳过: Every trading week must have a review, even if no trades were taken.
4. 归因诚实 (Honest Attribution): System-external errors must be labeled as such. Do not rationalize emotional decisions as "rule-adherent."
5. 规则迭代闭环 (Rule Iteration Loop): Rule gaps identified in weekly review → propose rule update → backtest with historical_sop_replay.py → if validated, adopt in next version.

Common Pitfalls

1. Skipping the fundamental check: A technically perfect setup on a company with structural problems is still a bad trade.
2. Relaxing R:R threshold: 3:1 is the minimum. A "2.5:1 almost there" trade is a reject. The math requires >3:1 to survive a 40% win rate.
3. Not recording untriggered signals: The best learning often comes from "why didn't I take that trade?" — record it.
4. Confusing luck with skill in review: A rule-adherent win doesn't mean the rule is perfect. A rule-adherent loss doesn't mean the rule is broken. Look at the distribution over 20+ trades.
5. Adjusting stops mid-trade without re-evaluating R:R: Widening a stop to "give it room" changes the R:R. Re-compute and confirm it still exceeds 3:1.