marcus-investment-analyst

Marcus 投资分析技能 - 基于缠论+MACD+RSI 的 A 股投资策略。用于股票分析、回测、投资建议。触发词：分析股票、回测策略、投资建议、Marcus 策略。

Security Scan

The skill's code, instructions, and resource usage are consistent with an A‑share investment analysis and backtest tool; it fetches market data, computes indicators and writes/reads local SQLite databases as expected.

Purpose & Capability

Name/description (缠论+MACD+RSI 回测与分析) align with included scripts: chan theory analyzer, backtester, indicator fetcher, and a strategy using backtrader. Required capabilities (network data fetch via akshare, local DBs for history/indicators) are appropriate for the stated purpose.

Instruction Scope

Runtime instructions tell the agent to run the provided Python scripts and read packaged reference files — this matches the code. However the SKILL.md and scripts reference system paths outside the skill folder (/root/data/*.db and workspace memory paths) and will read/write local SQLite DBs and workspace files; that is within the domain of a backtester but is noteworthy because it modifies host filesystem under /root.

Install Mechanism

No install spec provided (instruction-only with source files). That lowers risk, but the scripts depend on third‑party Python packages (akshare, pandas, backtrader, etc.) which the environment must already provide; missing dependencies will cause runtime errors.

Credentials

The skill does not request environment variables or credentials. It uses network access (akshare) to fetch market data and writes to local DB files. No unexplained tokens/keys or unrelated service credentials are required.

Persistence & Privilege

always:false and no automated model disabling — normal. The scripts create/append SQLite DB files at /root/data/astock_indicators.db and read /root/data/astock_history.db; they will persist data on disk in host /root. This is proportional to a data-collection/backtest tool but you should be aware of the skill creating and modifying files in /root.

Assessment

This package appears to do what it says: fetch A‑share data (via akshare), compute indicators, run Chan‑theory analysis and run/backtest a strategy. Before installing or running: 1) ensure you trust the source and review the code (it will read/write SQLite DBs under /root/data and workspace paths); 2) run in an isolated environment or non‑privileged account (don’t run as root) to limit filesystem impact; 3) ensure required Python packages (akshare, pandas, backtrader, etc.) are installed in a controlled venv; 4) be aware it uses network access to fetch market data (akshare) — if you have data policies, check those; 5) back up any existing /root/data/astock_*.db files before running because the fetcher will append to or create those databases. If you want tighter containment, run the scripts after editing DB paths to a sandboxed directory.

Marcus 投资分析技能

版本: v6.0
策略类型: 缠论+MACD+RSI+ 追踪止损

---

🎯 何时使用

使用此技能当用户需要：

1. 分析 A 股股票（技术指标、缠论中枢、买卖点）
2. 执行策略回测（历史数据验证）
3. 获取投资建议（仓位配置、行业配置）
4. 下载/更新股票指标数据（MACD、RSI、KDJ）

---

📋 核心策略参数

策略名称：优化版 5-存储芯片专用

缠论中枢:
  周期：60 日均线
  区间：±8%

MACD:
  快线：12
  慢线：26
  信号：9

RSI:
  周期：6
  超卖：20
  超买：80

止损止盈:
  止损：5%
  止盈：15%
  追踪：8%

---

📈 买入条件 (需同时满足)

1. ✅ 价格在缠论中枢内 (60 日均线×0.92 ~ 60 日均线×1.08)
2. ✅ MACD 金叉 (DIF 上穿 DEA)
3. ✅ RSI6 < 20 (超卖)

---

📉 卖出条件 (满足任一)

1. ❌ 盈利 ≥ 15% (止盈)
2. ❌ 亏损 ≥ 5% (止损)
3. ❌ 从最高点回撤 ≥ 8% (追踪止损)
4. ❌ MACD 死叉 (DIF 下穿 DEA)
5. ❌ 价格跌破中枢下沿

---

🎯 股票池 (20 支)

核心配置 (55-60%)

股票	代码	行业	仓位
江波龙	301308	存储芯片	25-30%
兆易创新	603986	存储芯片	20-25%

卫星配置 (25-30%)

股票	代码	行业	仓位
东山精密	002384	消费电子	10-15%
金风科技	002202	风电	10%
四方精创	300468	软件	10%
云天化	600096	化工	8%
宝丰能源	600989	化工	8%
万向钱潮	000559	汽配	5%
中国铝业	601600	有色	5%

回避

• 广汇能源 (600256)
• 创新医疗板块

---

🚀 使用脚本

1. 运行策略回测

cd /root/.openclaw/workspace/skills/marcus-investment-analyst/scripts
python3 marcus_ultimate_optimized_strategy.py

输出: 20 支股票的回测结果，包含收益、夏普比率、胜率

2. 缠论分析单只股票

python3 marcus_chan_theory.py <股票代码>
# 示例：python3 marcus_chan_theory.py 301308

输出: 缠论中枢、背驰、买卖点分析

3. 缠论回测

python3 marcus_backtest_chan.py <股票代码>
# 示例：python3 marcus_backtest_chan.py 301308

输出: 缠论策略历史回测结果

4. 下载指标数据

python3 data_indicator_fetcher.py

输出: 更新 MACD/RSI/KDJ 指标到数据库

---

📊 回测表现 (2023-2026)

指标	数值
平均年化收益	21.96%
夏普比率	0.56
胜率	75.0%
用户股票平均	28.77%

---

📁 文件结构

marcus-investment-analyst/
├── SKILL.md                    # 本文件
├── scripts/
│   ├── marcus_ultimate_optimized_strategy.py  # 主策略
│   ├── marcus_chan_theory.py                  # 缠论分析
│   ├── marcus_backtest_chan.py                # 缠论回测
│   └── data_indicator_fetcher.py              # 指标下载
├── references/
│   ├── strategy.md                            # 策略文档
│   └── optimization_report.md                 # 优化报告
└── assets/
    └── backtest_data.json                     # 回测数据

---

💡 使用示例

示例 1: 分析股票

用户: "分析一下江波龙"

操作:

1. 读取 references/strategy.md 获取股票池信息
2. 运行 python3 marcus_chan_theory.py 301308
3. 返回缠论分析结果

示例 2: 回测策略

用户: "回测一下 Marcus 策略"

操作:

1. 运行 python3 marcus_ultimate_optimized_strategy.py
2. 读取 assets/backtest_data.json 获取历史数据
3. 返回回测结果和投资建议

示例 3: 投资建议

用户: "有什么投资建议？"

操作:

1. 读取 references/optimization_report.md
2. 返回行业配置建议和个股推荐

---

⚠️ 风险提示

1. 历史业绩不代表未来
2. 存储芯片集中度高，建议分散配置
3. 市场系统性风险无法避免
4. 流动性风险：小盘股可能无法及时买卖

---

📞 数据位置

数据库：
/root/data/astock_history.db    # K 线数据 (154MB)
/root/data/astock_indicators.db # 指标数据 (283MB)

回测数据：
/root/.openclaw/workspace/memory/stock-analysis/终极优化策略回测_20260312_1629.json

---

最后更新: 2026-03-12
状态: ✅ 实盘就绪