選股秘笈15 分鐘

Advanced Stock Screening Techniques: A Comprehensive Multi-Factor Framework Combining Fundamentals, Technicals, and Quantitative Factors

Explore how to combine fundamental, technical, and sentiment factors to build a repeatable multi-factor scoring system, using normalization, weighting, and backtest validation to achieve robust excess returns across different market regimes.

Algo Lab Team發布於 2026-05-08 16:00

重點摘要

Multi-factor stock analysis is the process of combining signals from fundamental indicators, technical indicators, and sentiment/alternative data into a single, repeatable decision framework. According to StockAlpha.ai (2026), combining factors from different domains reduces reliance on any single data source and helps manage strategy risk across different market regimes. Fundamental factors should measure economic durability and valuation (ROIC, EBITDA margin, Net Debt/EBITDA), technical factors should confirm price structure and momentum (200-day moving average, 12-week momentum, 20-day ATR volatility), while sentiment factors measure investor attention and short-term crowd behavior (social media sentiment, options flow, short interest).

Multi-Factor Stock Analysis: From Single Dimension to Comprehensive Framework

Traditional stock screening methods often rely on a single dimension — either fundamentals (P/E ratio, dividend yield) or technicals (moving averages, RSI). However, according to StockAlpha.ai (2026), "multi-factor stock analysis is the process of combining signals from fundamental indicators, technical indicators, and sentiment/alternative data into a single, repeatable decision framework." This hybrid approach reduces reliance on any single data source and helps manage strategy risk across different market regimes. You can also experience the power of multi-factor screening through our platform's Stock Radar.

Why a Multi-Factor Approach?

Single-Factor ApproachMulti-Factor Framework
Vulnerable to specific market regimesRobust across market regimes
Rapid signal decayComplementary factors extend signal life
Hard to explain performanceAttributable to specific factor contributions
High volatilityBetter risk-adjusted returns

According to fffinstill (2026) research, "94 fundamental, factor, and forward-looking filters combined with a complete point-in-time backtesting engine, a time machine that replays historical snapshots, AI-driven analysis, and portfolio-level diagnostics far beyond simple stock lists." This shows that modern stock selection has shifted from single indicators to multi-dimensional integration.

Building a Three-Layer Factor Architecture

Layer 1: Fundamental Factors (40% Weight)

Fundamental factors should measure economic durability and valuation. According to StockAlpha.ai (2026), typical selections include:

Profitability Metrics:

  • Return on Invested Capital (ROIC): Measures capital efficiency
  • EBIT Margin: Core profitability excluding financing structure
  • Free Cash Flow Conversion Rate: Ability to convert profits into cash

Earnings Quality Metrics:

  • Accruals: Identifies earnings management risk
  • Cash Conversion Rate: Avoids landmine stocks with book profits but deteriorating cash flow

Balance Sheet Strength:

  • Net Debt/EBITDA: Debt pressure assessment
  • Current Ratio and Quick Ratio: Short-term solvency
  • Interest Coverage Ratio: Long-term solvency

Each factor should have a clear directional expectation. For example, higher ROIC and EBITDA margins are expected to bring better future returns, while higher Net Debt/EBITDA is expected to underperform.

Layer 2: Technical Factors (40% Weight)

Technical factors should confirm price structure and momentum. Useful inputs suggested by StockAlpha.ai include:

Trend Confirmation:

  • 200-day Moving Average: Long-term trend determination
  • Price Above 200-day MA: Basic condition for bullish alignment

Momentum Indicators:

  • 12-week or 6-month Total Return: Medium-term momentum assessment
  • Volatility-Adjusted Momentum: Momentum divided by realized volatility or information ratio style metrics

Volatility and Volume:

  • 20-day ATR (Average True Range): Risk measurement
  • OBV (On-Balance Volume) or VWAP Deviation: Capital flow confirmation

According to fffinstill (2026) empirical evidence, "use trend filters (e.g., price > 200-day MA) and shorter-term momentum (e.g., 3-6 month total return) for multi-timeframe confirmation." Additionally, "use percentile rank cross-universe normalization to reduce size sensitivity between high-priced and low-priced stocks."

Layer 3: Sentiment and Alternative Data Factors (20% Weight)

Sentiment data measures investor attention, news tone, and short-term crowd behavior. Common sources include:

Social Media Sentiment Scores:

  • Sentiment analysis of brand mentions on Twitter/X, Reddit, Facebook
  • LLM-based qualitative sentiment parsing of earnings calls

Options Flow:

  • Unusual call volumes
  • Put/Call Ratio

Short Interest:

  • Short interest balance
  • Days to cover

News Sentiment Algorithms: According to the 2025 AAII (American Association of Individual Investors) review, "Wanger's investment approach focuses on identifying established small-cap companies that dominate specific market niches and are positioned to benefit from long-term themes expected to last at least four to five years." This combination of fundamental strength with long-term trend confirmation demonstrates the practical value of multi-factor thinking.

Factor Normalization and Weighting: Key Technical Details

Normalization Methods: Z-Scores vs. Percentile Ranks

Combining factors requires normalization, weighting, and aggregation. According to StockAlpha.ai, two common normalization methods are z-scores and ranks/percentiles, each with trade-offs:

MethodAdvantagesDisadvantagesBest For
Z-ScoresPreserves distance information, identifies extremesSensitive to outliersWhen precise deviation measurement is needed
Percentile RanksRobust to outliers, eliminates scale effectsLoses distance informationIntegrating factors with different units

Weighting Strategies: Three Main Approaches

According to StockAlpha.ai research, three main weighting methods:

  1. Equal Weighting: Equal weights across factor groups (fundamental, technical, sentiment) — robust and easy to interpret
  2. Signal Quality Weighting: Uses in-sample predictive strength (IC, Information Coefficient) to weight factors proportionally by historical performance
  3. Risk-Based Weighting: Scales factor contributions by their volatility or correlation, reducing concentration in any single signal

Composite Score Formula Example

According to fffinstill (2026)'s proprietary scoring model:

Composite Score = 0.4 × Fundamental Score + 0.4 × Technical Score + 0.2 × Sentiment Score

Where:

  • Fundamental Score: Piotroski F-Score (0-9) + Altman Z-Score (0-100) + Custom Profitability Score
  • Technical Score: Momentum Percentile + Trend Confirmation Binary + Volatility-Adjusted Score
  • Sentiment Score: Social Sentiment Percentile + Options Flow Z-Score + Inverse Short Interest

Rigorous Backtesting and Validation Framework

Avoiding Look-Ahead Bias

According to StockAlpha.ai, "backtest across multiple market regimes (bull, bear, sideways) to understand factor behavior and drawdown characteristics. Use walk-forward optimization and out-of-sample testing to avoid over-optimization."

Key Validation Steps:

  1. Information Coefficient (IC) Calculation: Calculate IC for each factor and combination on a rolling window to measure predictive stability
  2. Turnover and Transaction Cost Analysis: Multi-factor models with frequent re-ranking may generate high transaction costs
  3. Stress Testing: Extreme volatility, liquidity shocks, and sector crashes — observe model behavior

Walk-Forward Validation: Simulating Real Deployment

According to fffinstill (2026)'s robustness analysis suite:

  • Walk-Forward Test: Divides history into expanding in-sample windows and 12-month out-of-sample windows. Reports average IS CAGR, OOS CAGR, performance decay, and OOS win rate. Low decay means the strategy is generalizable.
  • Sensitivity Analysis: Perturb each numerical filter by ±20% and re-run the backtest. If performance collapses with small changes, the strategy is fragile.
  • Sub-Period Analysis: Reports returns for each sub-period. Consistent performance across sub-periods is a strong signal.

Practical Risk Controls

According to StockAlpha.ai recommendations:

  1. Stop-loss or Trend Failure Exit Rules: e.g., exit if price falls below 50-day EMA by X%
  2. Maximum Position Limit: No single position exceeds 2-3% of portfolio
  3. Sector Cap: Single sector exposure does not exceed 20-30%
  4. Dynamic Rebalancing: Adjust rebalancing frequency based on factor IC decay speed

Real-World Screening Strategy Case Studies

Value and Momentum Composite Strategy

According to fffinstill (2026) empirical case studies:

  • Piotroski Value Strategy: F-Score ≥ 7, P/B ≤ 1.5, equal weight, quarterly rebalance
  • Momentum Quality Strategy: 52-week return ≥ 20%, ROIC ≥ 12%, Sharpe Ratio ≥ 1, inverse volatility weighting
  • Safe Dividend Strategy: Yield ≥ 3%, score-weighted, quarterly rebalance

Backtest Results (fffinstill data):

  • Piotroski Value: 16.8% risk-adjusted gain since 1998, despite risk index as high as 2.04
  • Momentum Quality: Sharpe Ratio 1.8+, max drawdown < 15%
  • Safe Dividend: Correlation with market < 0.3, significant defensive characteristics

AAII Screening Strategy Annual Review

According to the 2025 AAII annual review, the best-performing strategies include:

  1. O'Shaughnessy Tiny Titans: Small-cap momentum strategy, 16.8% risk-adjusted gain
  2. Wanger's Zebras: Identifies small-cap companies dominating specific market niches, avoiding IPOs and startups
  3. Quality Compounder: ROIC ≥ 15%, Health Score ≥ 70, Revenue Growth ≥ 10%, score-weighted

According to AAII's analysis, "most stocks are positively correlated with the overall market: when the market rises, most stocks follow; when the market falls, their values decline. By analyzing stock screening performance in recent bull and bear markets, we can evaluate a strategy's ability to profit in uptrends and limit losses during downturns."

Modern Tools and AI Enhancement

fffinstill Screener: The Power of 94 Filters

According to fffinstill (2026), their screener includes:

  • 94 Fundamental, Factor, and Forward-Looking Filters: Piotroski F-Score, Altman Z-Score, M-Score, FF Score, Momentum
  • Time Machine: Replays historical snapshots to avoid look-ahead bias
  • AI-Driven Analysis: Automatically identifies factor decay and recalibration timing
  • Portfolio-Level Diagnostics: Risk analysis beyond simple stock lists

Proprietary Scoring Models:

  • Trajectory Conviction Score: 0-100 score measuring multi-quarter fundamental momentum
  • Fundamental Beta: Earnings-based beta (not price-based), measuring fundamental cyclicality
  • Composite Robustness Score: Combines walk-forward, sensitivity, sub-period, and complexity penalties — 75+ is robust, 40- is likely over-optimized

Stock Rover Advanced Screening

According to Stock Rover (2026):

  • Rank Screening: Ranks targets by custom weights rather than just pass/fail
  • Three Power Indicators: Altman Z-Score (0-100), Piotroski F-Score (0-9), Stock Rover Health Score (0-100)
  • GARP – Premium Plus: Seeks "Growth at a Reasonable Price," prioritizing attractive valuation ratios (e.g., PEG) while enforcing consistent historical earnings growth and positive safety margins
  • Value with Growth: Hunts for the "sweet spot" of investing — low valuation multiples (e.g., EV/EBITDA) but high expected earnings growth

Conclusion: From Single Indicators to Systematic Process

Multi-factor stock selection is not simply stacking indicators — it is about building a repeatable, explainable, and adaptable systematic process. According to StockAlpha.ai's summary, the key steps include:

  1. Design factors from different domains with clear economic rationale: Fundamentals, technicals, and sentiment each have their role
  2. Normalize heterogeneous inputs and choose weights that balance signal quality and diversification
  3. Backtest across multiple regimes using walk-forward validation and control for look-ahead bias
  4. Implement practical filters: Quality/valuation first, trend confirmation second, sentiment as timing or risk overlay
  5. Monitor factor decay and turnover, and adapt thresholds rather than fixed rules

According to CMS Financial's 2026 empirical study, "long-term outperformance rarely comes from pure fundamentals or perfect timing. Instead, the most resilient alpha emerges where fundamental quality and technical momentum converge at precise inflection points." This dual-lens approach tilts the odds decisively in the trader's favor. Visit our Strategy Center to explore more systematic stock selection solutions.

Actionable Next Steps:

  1. Choose a universe (e.g., S&P 500 or Russell 2000)
  2. Define 6-10 sub-factors with documented rules
  3. Normalize using percentiles or z-scores
  4. Backtest with realistic costs
  5. Implement risk controls
  6. Iteratively monitor factor IC and recalibrate, rather than constantly adding new signals

References:

  1. StockAlpha.ai (2026). "Multi-Factor Stock Analysis: Blending Fundamental, Technical and Sentiment Signals." Alpha Learning.
  2. fffinstill Research (2026). "Inside the fffinstill Screener: Point-in-Time Backtesting, Time Machine, AI Analysis & 94 Filters."
  3. AAII (2026). "2025 Review of AAII Stock Screens: Wanger's Zebras Break From the Herd."
  4. CMS Financial (2025). "A Dual-Lens Stock-Picking Framework: Fundamental Quality Meets Technical Timing." LinkedIn Pulse.
  5. Stock Rover (2026). "Advanced Screening with Stock Rover."
  6. Diligize (2025). "Quantitative Analysis Stock Market: A Practical Playbook for 2025."
  7. Investopedia. "Factor Investing." https://www.investopedia.com/terms/f/factor-investing.asp
  8. NASDAQ. "Stock Screening Strategies." https://www.nasdaq.com/
#Advanced Stock Screening#Multi-Factor Model#Fundamental Analysis#Technical Analysis#Quantitative Screening#advanced stock screening#multi-factor stock analysis#fundamental technical quantitative#stock screener 2025#factor investing strategy#Piotroski F-Score#Altman Z-Score#composite scoring system#walk-forward validation#stock selection framework#quantitative screening tools

上一篇

Alternative Data in Quantitative Trading: Practical Integration of Satellite Imagery, Credit Card Data, and Web Scraping

想要每日接收高勝率信號?

訂閱 VIP 會員,每日 TOP 20 信號推送,形態識別 + AI 選股,助你做出明智投資決策。

延伸閱讀

文章相關問題