| 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:
- 确认时间 (Confirmation Time): YYYY-MM-DD HH:MM
- 确认前提 (Confirmation Premise): What signals were checked (L1 state, L2 stage, L3 scores, fundamental check)
- 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
- 盘后执行 (Post-Market Only): Trade plan generation happens after market close. No real-time plan adjustment during trading hours.
- 计划盈亏比硬性门槛: R:R < 3:1 → auto-reject, do not pass to confirmation.
- 每周复盘不可跳过: Every trading week must have a review, even if no trades were taken.
- 归因诚实 (Honest Attribution): System-external errors must be labeled as such. Do not rationalize emotional decisions as "rule-adherent."
- 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
- Skipping the fundamental check: A technically perfect setup on a company with structural problems is still a bad trade.
- 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.
- Not recording untriggered signals: The best learning often comes from "why didn't I take that trade?" — record it.
- 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.
- 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.
