SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsquant-factor-tracker

quant-factor-tracker

Data & analytics

View repository

Updated 10-15 19:55

About this skill

Quant Factor Tracker

Overview

This skill is designed specifically forsecurities financial engineering researchers, aiming to provide systematicquantitative factor tracking and effectiveness evaluationBy automatically obtaining or receiving cross-sectional data, calculating factor IC/IR and grouped excess returns, performing monotonicity and decay tests, marking factor strength status and failure warnings, and generating standardized Markdown-format factor tracking reports, it assists in multi-factor model construction and strategy iteration.

All data queries must use the gildata-aidata service。

Core Analysis Framework

1. Factor Effectiveness

  • Rank IC (Information Coefficient): The rank correlation coefficient between factor values and next-period returns.
    • IC Mean: Measures the strength of the factor's predictive ability. Typically |IC| > 0.03 is considered effective.
    • ICIR (IC / std(IC)): Measures the stability of the factor's predictive ability. Typically ICIR > 0.5 is considered excellent.
    • IC win rate: The proportion of periods with IC > 0.
  • Grouped Returns:
    • Stocks are divided into 5 or 10 groups by factor value.
    • Long-Short Return: The return difference between Group 1 (highest) - Group 5 (lowest).
    • Monotonicity: Whether the returns of each group are strictly monotonically increasing/decreasing.

2. Factor Decay & Turnover

  • IC decay: Calculate the IC of factor values with T+1, T+2, ... T+N period returns, and observe the decay speed of predictive ability over time.
  • Factor turnover: The magnitude of change in factor value rankings, assessing the erosion of factor returns by trading costs.

3. Status Monitoring

  • 🟢 Strong factor: ICIR > 0.5, significant long-short return, good monotonicity.
  • 🟡 Oscillating/Watch: The mean IC is acceptable but volatile, or the long-short return is unstable.
  • 🔴 Failure/Warning: The mean IC is close to 0, ICIR < 0.2, or a reverse signal appears (IC remains persistently negative).

Data Collection Guide (gildata-aidata tool mapping)

All data must be obtained through the gildata-aidata service, and the specific tool calls are as follows:

1. Factor Cross-Sectional Data Acquisition

According to the factor type, call the corresponding tool to obtain cross-sectional data for the whole market or a specified stock pool:

  • Value factors (Value): PE(TTM), PB(LF), PS(TTM), dividend yield. 
    finx gildata-aidata StockValueAnalysis --body '{"query": "查询沪深300成分股最新PE、PB、PS、股息率"}'
  • Growth/Quality factors (Growth/Quality): ROE, net profit growth rate, revenue growth rate, gross margin. 
    finx gildata-aidata FinancialAnalysis --body '{"query": "查询全市场A股最新ROE、净利润同比增长率、毛利率"}'
  • Momentum/Reversal factors (Momentum/Reversal): Past N-day return, volatility. 
    finx gildata-aidata StockRangeQuotation --body '{"query": "查询全市场A股过去20日、60日涨跌幅"}'
  • Volatility/Risk factors (Volatility/Beta): Beta coefficient, specificity, historical volatility. 
    finx gildata-aidata StockRiskAnalysis --body '{"query": "查询全市场A股最新Beta系数、波动率"}'
  • Fund Flow Factor (Fund Flow): Main force net inflow ratio, Northbound capital holdings changes. 
    finx gildata-aidata AStockCashFlow --body '{"query": "查询全市场A股近5日主力资金净流入占比"}'

2. Obtaining Next-Period Returns (for calculating IC)

  • Obtain the next-period return after the factor exposure date (e.g., T+1 day or T+5 day): 
    finx gildata-aidata StockDailyQuote --body '{"query": "查询指定日期全市场A股涨跌幅"}'

3. Stock Pool Definition

  • Obtain constituent stocks of a specific index as the stock pool: 
    finx gildata-aidata IndexConstituentStocks --body '{"query": "查询沪深300最新成分股列表及权重"}'

Execution Flow

  1. Data Collection:
    • Use gildata-aidata to obtain the factor cross-sectional data (Factor Exposure) of the specified stock pool.
    • Obtain the next-period return (Return) after the cross-sectional date.
  2. Factor Calculation:
    • Data Cleaning: Remove ST stocks, suspended stocks, new stocks, and financial stocks (depending on the strategy).
    • Outlier Removal and Standardization: Perform MAD outlier removal and Z-Score standardization on the factors.
    • Industry and Market Cap Neutralization (optional): Remove the influence of industry and market cap factors.
  3. Performance Evaluation:
    • Calculate the Rank IC series.
    • Calculate group returns (Top/Bottom group returns and long-short returns).
    • Execute monotonicity test (Spearman correlation coefficient).
  4. Report Generation:
    • Generate a factor tracking and evaluation report in Markdown format.

Report Template (Markdown)

# 📊 量化因子跟踪与评估报告 | [日期]

> **数据来源**: gildata-aidata 专业金融数据服务
> **测试区间**: [起始日期] 至 [结束日期]
> **股票池**: [如：沪深300 / 中证500 / 全A]

---

## 一、 因子有效性概览 (Factor Overview)

| 因子名称 | IC 均值 | ICIR | IC 胜率 | 多空年化 | 单调性 | 状态 |
|----------|---------|------|---------|----------|--------|------|
| EP_TTM   | 0.045   | 0.65 | 62%     | +8.5%    | ✅ 优   | 🟢 强势 |
| ROE_TTM  | 0.032   | 0.45 | 58%     | +5.2%    | ⚠️ 一般 | 🟡 观察 |
| Vol_20D  | -0.015  | -0.20| 40%     | -3.1%    | ❌ 差   | 🔴 失效 |
| ...      | ...     | ...  | ...     | ...      | ...    | ...  |

---

## 二、 核心因子深度分析 (Deep Dive)

### 2.1 [因子名称] (如：EP_TTM)
*   **IC 序列分析**:
    *   近一月 IC 均值：**0.045** (环比上升/下降)。
    *   当前 ICIR：**0.65** (稳定性优秀)。
    *   *解读*: 价值因子近期表现稳健，市场风格偏向低估值。
*   **分组收益分析**:
    *   **Top 组年化**: +12.5%
    *   **Bottom 组年化**: +4.0%
    *   **多空收益**: +8.5% (单调性良好，第 1 组显著跑赢第 5 组)。
*   **衰减分析**:
    *   T+1 IC: 0.045
    *   T+5 IC: 0.030
    *   T+20 IC: 0.010
    *   *结论*: 预测能力随时间衰减较快，适合中低频调仓（月度）。

---

## 三、 因子状态预警 (Alerts)

*   **🟢 强势因子推荐**:
    *   **[因子 1]**: ICIR 持续大于 0.5，建议加大配置权重。
    *   **[因子 2]**: 近期出现拐点，IC 由负转正，关注风格切换。
*   **🔴 失效/风险预警**:
    *   **[因子 3]**: 近 3 个月 IC 均值持续为负，多空收益倒挂，建议剔除或反向使用。
    *   **[因子 4]**: 拥挤度上升，换手率激增，需警惕交易成本侵蚀收益。

---

## 四、 策略建议 (Strategy Implication)

1.  **模型调整**: 建议降低 [失效因子] 权重，提升 [强势因子] 权重。
2.  **风格暴露**: 当前模型在 [成长/价值/小盘] 风格上有显著暴露，注意风格漂移风险。
3.  **下期展望**: 基于宏观环境，预期 [某类因子] 将在下阶段占优。

---

*免责声明：本报告基于 gildata-aidata 数据和量化模型生成，仅供参考，不构成投资建议。*

Notes

  • Data Source Uniqueness: All factor data and return data must be obtained through the gildata-aidata service.
  • Neutralization Processing: When evaluating a single factor, be sure to consider the influence of industry and market cap; neutralization processing is recommended to extract pure Alpha.
  • Out-of-Sample Testing: Factor evaluation should include both in-sample and out-of-sample testing to avoid overfitting.
  • Dynamic Tracking: Factor effectiveness changes with market cycles, so the tracking report needs to be updated regularly (weekly/monthly).