mcpField
U
us-market-bubble-detector

us-market-bubble-detector

github.com/tradermonty其他3.4k3.4k 次安装2 天前更新
代码仓库

关于 us-market-bubble-detector

用于评估美国市场泡沫风险的投资分析技能,面向需要制定仓位与风险管理判断的市场参与者。它先收集量化指标并按阈值评分,再以有限的定性调整得出风险阶段。

主要功能

  1. 01收集市场结构与杠杆数据
  2. 02按六项指标量化评分
  3. 03定性调整上限为 3 分
  4. 04划分五个风险阶段
  5. 05提供仓位与做空指引

适用场景

  1. 01评估市场是否处于泡沫阶段
  2. 02根据风险分数调整投资敞口
  3. 03检查做空条件是否满足

安装

复制下面的命令,在终端中运行。

$npx skills add https://github.com/tradermonty/claude-trading-skills --skill us-market-bubble-detector

安装后在 Claude Code 中输入/us-market-bubble-detector确认已加载。

包含的文件

除 SKILL.md 外,安装时一起下载的参考资料和脚本。

CHANGELOG.md
requirements.txt
references/
bubble_framework.md
historical_cases.md
implementation_guide.md
quick_reference_en.md
quick_reference.md
scripts/
bubble_scorer.py
scripts/tests/
test_skill_contract.py

详细介绍

美国市场泡沫检测技能(修订版 v2.1)

v2.1 的主要修订

相较于 v2.0 的关键变更:

  1. ✅ 强制收集量化数据——使用测量值,而非印象或猜测
  2. ✅ 明确设置阈值——为每项指标设定具体数值标准
  3. ✅ 两阶段评估流程——量化评估 → 定性调整(严格按顺序)
  4. ✅ 更严格的定性标准——最高 +3 分(从 +5 分下调),且需有可测量的证据
  5. ✅ 防止确认偏误——提供明确的检查清单,避免过度评分
  6. ✅ 更细致的风险阶段——新增“风险升高”阶段(8–9 分),以便进行更细致的风险管理

何时使用此技能

在以下情况下使用此技能:

英语:

  • 用户询问“市场是否处于泡沫中?”或“我们是否正处于泡沫中?”
  • 用户寻求有关获利了结、新建仓时机或卖空决策的建议
  • 用户报告社会现象(非投资者入场、媒体狂热、IPO 激增)
  • 用户提到“这次不一样”或“革命性技术”等叙事开始成为主流
  • 用户咨询现有持仓的风险管理

日语:

  • 用户询问“现在的市场是不是泡沫?”
  • 用户寻求投资获利了结、新建仓或卖空时机的判断
  • 用户观察到社会现象(非投资者入场、媒体过热、IPO 泛滥)并表达担忧
  • 用户报告“这次不一样”“革命性技术”等叙事成为主流
  • 用户咨询如何管理持仓风险

评估流程(严格按顺序)

阶段 1:强制收集量化数据

重要:开始评估前,务必收集以下数据

1.1 市场结构数据(优先级最高)
□ Put/Call Ratio (CBOE Equity P/C)
  - Source: CBOE DataShop or web_search "CBOE put call ratio"
  - Collect: 5-day moving average

□ VIX (Fear Index)
  - Source: Yahoo Finance ^VIX or web_search "VIX current"
  - Collect: Current value + percentile over past 3 months

□ Volatility Indicators
  - 21-day realized volatility
  - Historical position of VIX (determine if in bottom 10th percentile)
1.2 杠杆与持仓数据
□ FINRA Margin Debt Balance
  - Source: web_search "FINRA margin debt latest"
  - Collect: Latest month + Year-over-Year % change

□ Breadth (Market Participation)
  - % of S&P 500 stocks above 50-day MA
  - Source: web_search "S&P 500 breadth 50 day moving average"
1.3 IPO 与新发行数据
□ IPO Count & First-Day Performance
  - Source: Renaissance Capital IPO or web_search "IPO market 2025"
  - Collect: Quarterly count + median first-day return

⚠️ 重要:未完成阶段 1 的数据收集,不得继续评估


阶段 2:量化评估(量化评分)

根据收集的数据,按照以下标准进行机械评分:

指标 1:看跌/看涨期权比率(市场情绪)
Scoring Criteria:
- 2 points: P/C < 0.70 (excessive optimism, call-heavy)
- 1 point: P/C 0.70-0.85 (slightly optimistic)
- 0 points: P/C > 0.85 (healthy caution)

Rationale: P/C < 0.7 is historically characteristic of bubble periods
指标 2:波动率受抑与新高
Scoring Criteria:
- 2 points: VIX < 12 AND major index within 5% of 52-week high
- 1 point: VIX 12-15 AND near highs
- 0 points: VIX > 15 OR more than 10% from highs

Rationale: Extreme low volatility + highs indicates excessive complacency
指标 3:杠杆(保证金债务余额)
Scoring Criteria:
- 2 points: YoY +20% or more AND all-time high
- 1 point: YoY +10-20%
- 0 points: YoY +10% or less OR negative

Rationale: Rapid leverage increase is a bubble precursor
指标 4:IPO 市场过热
Scoring Criteria:
- 2 points: Quarterly IPO count > 2x 5-year average AND median first-day return +20%+
- 1 point: Quarterly IPO count > 1.5x 5-year average
- 0 points: Normal levels

Rationale: Poor-quality IPO flood is characteristic of late-stage bubbles
指标 5:市场广度异常(领涨范围狭窄)
Scoring Criteria:
- 2 points: New high AND < 45% of stocks above 50DMA (narrow leadership)
- 1 point: 45-60% above 50DMA (somewhat narrow)
- 0 points: > 60% above 50DMA (healthy breadth)

Rationale: Rally driven by few stocks is fragile
指标 6:价格加速
Scoring Criteria:
- 2 points: Past 3-month return exceeds 95th percentile of past 10 years
- 1 point: Past 3-month return in 85-95th percentile of past 10 years
- 0 points: Below 85th percentile

Rationale: Rapid price acceleration is unsustainable

阶段 3:定性调整(修订版 v2.1)

上限:+3 分(低于 v2.0 的 +5 分)

⚠️ 防止确认偏误检查清单:

Before adding ANY qualitative points:
□ Do I have concrete, measurable data? (not impressions)
□ Would an independent observer reach the same conclusion?
□ Am I avoiding double-counting with Phase 2 scores?
□ Have I documented specific evidence with sources?
调整项 A:社会渗透度(0–1 分,标准严格)
+1 point: ALL THREE criteria must be met:
  ✓ Direct user report of non-investor recommendations
  ✓ Specific examples with names/dates/conversations
  ✓ Multiple independent sources (minimum 3)

+0 points: Any criteria missing

⚠️ INVALID EXAMPLES:
- "AI narrative is prevalent" (unmeasurable)
- "I saw articles about retail investors" (not direct report)
- "Everyone is talking about stocks" (vague, unverified)

✅ VALID EXAMPLE:
"My barber asked about NVDA (Nov 1), dentist mentioned AI stocks (Nov 2),
Uber driver discussed crypto (Nov 3)"
调整项 B:媒体/搜索趋势(0–1 分,必须进行测量)
+1 point: BOTH criteria must be met:
  ✓ Google Trends showing 5x+ YoY increase (measured)
  ✓ Mainstream coverage confirmed (Time covers, TV specials with dates)

+0 points: Search trends <5x OR no mainstream coverage

⚠️ CRITICAL: "Elevated narrative" without data = +0 points

HOW TO VERIFY:
1. Search "[topic] Google Trends 2025" and document numbers
2. Search "[topic] Time magazine cover" for specific dates
3. Search "[topic] CNBC special" for episode confirmation

✅ VALID EXAMPLE:
"Google Trends: 'AI stocks' at 780 (baseline 150 = 5.2x).
Time cover 'AI Revolution' (Oct 15, 2025).
CNBC 'AI Investment Special' (3 episodes Oct 2025)."

⚠️ INVALID EXAMPLE:
"AI/technology narrative seems elevated" (unmeasurable)
调整项 C:估值脱节(0–1 分,避免重复计分)
+1 point: ALL criteria must be met:
  ✓ P/E >25 (if NOT already counted in Phase 2 quantitative)
  ✓ Fundamentals explicitly ignored in mainstream discourse
  ✓ "This time is different" documented in major media

+0 points: P/E <25 OR fundamentals support valuations

⚠️ SELF-CHECK QUESTIONS (if ANY is YES, score = 0):
- Is P/E already in Phase 2 quantitative scoring?
- Do companies have real earnings supporting valuations?
- Is the narrative backed by fundamental improvements?

✅ VALID EXAMPLE for +1:
"S&P P/E = 35x (vs historical 18x).
CNBC article: 'Earnings don't matter in AI era' (Oct 2025).
Bloomberg: 'Traditional metrics obsolete' (Nov 2025)."

⚠️ INVALID EXAMPLE:
"P/E 30.8 but companies have real earnings and AI has fundamental backing"
(fundamentals support = +0 points)

阶段 3 总计:最高 +3 分


阶段 4:最终判断(修订版 v2.1)
Final Score = Phase 2 Total (0-12 points) + Phase 3 Adjustment (0 to +3 points)
Range: 0 to 15 points

Judgment Criteria (with Risk Budget):
- 0-4 points: Normal (Risk Budget: 100%)
- 5-7 points: Caution (Risk Budget: 70-80%)
- 8-9 points: Elevated Risk (Risk Budget: 50-70%) ⚠️ NEW in v2.1
- 10-12 points: Euphoria (Risk Budget: 40-50%)
- 13-15 points: Critical (Risk Budget: 20-30%)

v2.1 的关键变更:

  • 新增“风险升高”阶段(8–9 分),以便更细致地调整仓位
  • 9 分不再属于极端防御区间(此前风险预算为 40%)
  • 现在在 8–9 分时允许 50–70% 的风险预算
  • 从“谨慎”阶段到“亢奋”阶段的过渡更加渐进

数据来源(必需)

美国市场
日本市场

实施检查清单

使用时请核实以下内容:

□ Have you collected all Phase 1 data?
□ Did you apply each indicator's threshold mechanically?
□ Did you keep qualitative evaluation within +3 point limit?
□ Are you NOT assigning points based on news article impressions?
□ Does your final score align with other quantitative frameworks?

重要原则(修订版)

1. 数据 > 印象

不要理会缺乏量化数据的“许多新闻报道”或“专家持谨慎态度”。

2. 严格顺序:量化 → 定性

始终按以下顺序评估:阶段 1(数据收集)→ 阶段 2(量化)→ 阶段 3(定性调整)。

3. 主观指标的上限

定性调整总计最多 +3 分,不能推翻量化评估。

4. “出租车司机”是象征性说法

没有非投资者的直接推荐,不要轻易认定已实现大规模渗透。


常见失误与解决方案(修订版)

失误 1:根据新闻报道进行评估

❌ “高市交易有很多报道” → 媒体饱和度 2 分 ✅ 核实 Google Trends 数值 → 根据测量值进行评估

失误 2:对专家评论反应过度

❌ “警告称市场过热” → 狂热区 ✅ 根据 Put/Call、VIX、保证金债务的测量值进行判断

失误 3:对价格上涨产生情绪化反应

❌ 1 天上涨 4.5% → 价格加速 2 分 ✅ 核实其在 10 年分布中的位置 → 客观评估

失误 4:仅根据估值进行判断

❌ P/E 为 17 → 估值脱节 2 分 ✅ 综合判断 P/E、对叙事的依赖程度及其他量化指标


按泡沫阶段采取的建议行动(修订版 v2.1)

正常(0-4 分)

风险预算:100%

  • 继续执行常规投资策略
  • 设置 ATR 2.0× 移动止损
  • 采用阶梯式止盈规则(上涨 +20% 时止盈 25%)

做空:不允许

  • 未满足综合条件(0/7 项)
警戒(5-7 分)

风险预算:70-80%

  • 开始部分止盈(减持 20-30%)
  • 将 ATR 收紧至 1.8×
  • 将新建仓位规模减少 50%

做空:不建议

  • 等待更明确的反转信号
风险升高(8-9 分)⚠️ v2.1 新增

风险预算:50-70%

  • 加大止盈力度(减持 30-50%)
  • 将 ATR 收紧至 1.6×
  • 新建仓位:高度精选,仅选择优质标的
  • 开始积累现金储备,为未来机会做准备

做空:谨慎考虑

  • 仅在确认至少 2/7 项综合条件后
  • 小规模试探性建仓(正常仓位的 10-15%)
  • 严格止损(ATR 2.0×)

新增阶段的理由: 该区间意味着需要提高警惕,但不必采取极端防御措施。 市场出现警示信号,但尚未濒临崩盘。 在保持优质仓位敞口的同时,积累灵活应对的空间。

狂热(10-12 分)

风险预算:40-50%

  • 加快阶梯式止盈(减持 50-60%)
  • 将 ATR 收紧至 1.5×
  • 除非出现大幅回调,否则不建立新的多头仓位

做空:积极考虑

  • 在确认至少 3/7 项综合条件后
  • 建立小规模仓位(正常仓位的 20-25%)
  • 仅采用风险明确的方式(期权、严格止损)
危急(13-15 分)

风险预算:20-30%

  • 大幅止盈或全面实施对冲
  • ATR 1.2× 或固定止损
  • 进入现金保全模式,为重大市场失衡做准备

做空:建议

  • 在确认至少 5/7 项综合条件后
  • 以小仓位分批入场,确认后逐步加仓
  • 严格止损(ATR 1.5× 或更高)
  • 考虑采用看跌期权以明确风险

做空的综合条件(7 项)

仅在确认以下条件中至少 3 项后考虑做空:

1. Weekly chart shows lower highs
2. Volume peaks out
3. Leverage indicators drop sharply (margin debt decline)
4. Media/search trends peak out
5. Weak stocks start to break down first
6. VIX surges (spike above 20)
7. Fed/policy shift signals

输出格式

评估报告结构(v2.1)
# [Market Name] Bubble Evaluation Report (Revised v2.1)

## Overall Assessment
- Final Score: X/15 points (v2.1: max reduced from 16)
- Phase: [Normal/Caution/Elevated Risk/Euphoria/Critical]
- Risk Level: [Low/Medium/Medium-High/High/Extremely High]
- Evaluation Date: YYYY-MM-DD

## Quantitative Evaluation (Phase 2)

| Indicator | Measured Value | Score | Rationale |
|-----------|----------------|-------|-----------|
| Put/Call | [value] | [0-2] | [reason] |
| VIX + Highs | [value] | [0-2] | [reason] |
| Margin YoY | [value] | [0-2] | [reason] |
| IPO Heat | [value] | [0-2] | [reason] |
| Breadth | [value] | [0-2] | [reason] |
| Price Accel | [value] | [0-2] | [reason] |

**Phase 2 Total: X/12 points**

## Qualitative Adjustment (Phase 3) - STRICT CRITERIA

**⚠️ Confirmation Bias Check:**
- [ ] All qualitative points have measurable evidence
- [ ] No double-counting with Phase 2
- [ ] Independent observer would agree

### A. Social Penetration (0-1 points)
- Evidence: [REQUIRED: Direct user reports with dates/names]
- Score: [+0 or +1]
- Justification: [Must meet ALL three criteria]

### B. Media/Search Trends (0-1 points)
- Google Trends Data: [REQUIRED: Measured numbers, YoY multiplier]
- Mainstream Coverage: [REQUIRED: Specific Time covers, TV specials with dates]
- Score: [+0 or +1]
- Justification: [Must have 5x+ search AND mainstream confirmation]

### C. Valuation Disconnect (0-1 points)
- P/E Ratio: [Current value]
- Fundamental Backing: [Yes/No - if Yes, score = 0]
- Narrative Analysis: [REQUIRED: Specific media quotes ignoring fundamentals]
- Score: [+0 or +1]
- Justification: [Must show fundamentals actively ignored]

**Phase 3 Total: +X/3 points (max reduced from +5 in v2.0)**

## Recommended Actions

**Risk Budget: X%** (Phase: [Normal/Caution/Elevated Risk/Euphoria/Critical])
- [Specific action 1]
- [Specific action 2]
- [Specific action 3]

**Short-Selling: [Not Allowed/Consider Cautiously/Active/Recommended]**
- Composite conditions: X/7 met
- Minimum required: [0/2/3/5] for current phase

## Key Changes in v2.1
- Stricter qualitative criteria (max +3, down from +5)
- Added "Elevated Risk" phase for 8-9 points
- Confirmation bias prevention checklist
- All qualitative points require measurable evidence

参考文档

references/implementation_guide.md(英文)- 建议首次使用
  • 包含强制数据收集的分步评估流程
  • NG 示例与 OK 示例
  • 自检质量标准(4 个等级)
  • 审查过程中的警示信号
  • 客观评估的最佳实践
references/bubble_framework.md(日文)
  • 详细的理论框架
  • Minsky/Kindleberger 模型说明
  • 行为心理学要素
references/historical_cases.md(日语)
  • 分析过去的泡沫案例
  • 互联网、加密货币、疫情泡沫
  • 提炼共性模式
references/quick_reference.md(日语)
references/quick_reference_en.md(英语)
  • 每日检查清单
  • 紧急三问评估
  • 快速评分指南
  • 关键数据来源
何时加载参考资料
  • 首次使用或需要详细指南: 加载 implementation_guide.md
  • 需要理论背景: 加载 bubble_framework.md
  • 需要历史背景: 加载 historical_cases.md
  • 日常操作: 加载 quick_reference.md(日语)或 quick_reference_en.md(英语)

总结:v2.1 修订要点

v2.0 的问题(2025 年 11 月发现):

  • 定性调整过于宽松(最高 +5)
  • “AI 叙事升温” → +1 分(无数据)
  • “P/E 30.8” → +1 分(与定量评估重复计分)
  • 结果:11/16 分——缺乏证据,过度看空

v2.1 的解决方案:

  • 收紧定性调整(最高 +3)
  • “AI 叙事升温” → 0 分(未量化)
  • “P/E 30.8,但 AI 有基本面支撑” → 0 分(有基本面支撑)
  • 结果:9/15 分——评估平衡、以数据为依据

主要改进:

  1. 防止确认偏差:添加定性分数前,必须明确完成检查清单
  2. 必须有可衡量的证据:没有具体数据(Google Trends、媒体报道)就不给分
  3. 防止重复计分:估值不得与第 2 阶段的定量评估重复
  4. 细化风险阶段:新增“风险升高”(8-9 分),以便更细致地调整仓位
  5. 平衡风险预算:9 分 = 50-70%(而非极端防守的 40%)

核心原则:

“我们信仰上帝;其他人必须拿出数据。”——W. Edwards Deming

2025 年的教训: 即使是以数据为依据的框架,也可能因主观的定性调整而失效。 v2.1 要求所有定性分数都必须有可衡量的证据。 独立观察者必须能够核实每一项调整。


版本历史:

  • v2.0(2025 年 10 月 27 日):强制收集定量数据
  • v2.1(2025 年 11 月 3 日):收紧定性标准、防止确认偏差、细化风险阶段

v2.1 修订原因: 防止因未经量化的“叙事”评估和重复计分而导致分数虚高。 确保所有泡沫风险评估均可独立核实,且不受确认偏差影响。

评论

支持 Markdown