作者：James P. O'Shaughnessy

What Works on Wall Street — Complete Implementation Specification

Based on James P. O'Shaughnessy, What Works on Wall Street (4th Edition, 2011)

The most comprehensive quantitative study of stock market factors ever published. O'Shaughnessy backtests dozens of single-factor and multi-factor strategies over 50+ years of data (1926–2009), identifying which factors actually predict stock returns and which are market folklore.

Overview
Methodology and Data
Value Factors
Growth Factors
Momentum and Relative Strength
Shareholder Yield
Small Cap vs Large Cap
Cornerstone Value Strategy
Cornerstone Growth Strategy
Multi-Factor Models
United Cornerstone Strategy
Sector Analysis
Risk and Drawdown Analysis
What Doesn't Work
Key Principles Summary

1. Overview

Core Thesis

O'Shaughnessy's central finding: simple, quantitative, factor-based strategies consistently beat both the market and professional fund managers over long periods. The reason most investors fail is not lack of information but lack of discipline — they cannot stick with a strategy through its inevitable periods of underperformance.

Key Conclusions

Value works. Cheap stocks (low P/E, P/S, P/B, P/CF, EV/EBITDA) outperform expensive stocks across all time periods tested.
Momentum works. Stocks with strong 6-12 month relative strength outperform weak momentum stocks.
Value + Momentum is the best combination. Buying cheap stocks with strong momentum produces the highest risk-adjusted returns.
Price-to-Sales (P/S) is the single best value factor. It is harder to manipulate than P/E and works across more market environments.
Shareholder yield (dividends + buybacks) is a powerful factor, often superior to dividend yield alone.
Small caps outperform large caps but with substantially higher volatility and worse drawdowns.
Consistency matters more than any single year's return. The best strategies are ones an investor can actually stick with.

"The central message of this book is simple: in the long run, strategies that make investment decisions based on the known, exposed, and tested performance of stocks beat those that rely on hope, stories, and subjective assessments."

2. Methodology and Data

2.1 Data Sources

Standard & Poor's Compustat database: point-in-time data from 1926 (some factors) and 1963 (most factors) through 2009.
Point-in-time construction: critical to avoid look-ahead bias. The data used for each year's screen is the data that was actually available to investors at that time.
All returns include dividends and are total returns.

2.2 Universe Definitions

O'Shaughnessy tests against several universes:

Universe	Definition	Approx Size
All Stocks	Market cap > $200M (inflation-adjusted)	~2,500 stocks
Large Stocks	Market cap > average market cap (top ~16% by market cap)	~500 stocks
Small Stocks	Market cap between $200M and market cap average	~2,000 stocks
Market Leaders	Market cap > $1B, earnings > average, strong financials	~350 stocks

2.3 Rebalancing

Annual rebalancing on December 31 of each year.
Portfolios are equal-weighted or market-cap-weighted depending on the test.
Portfolios typically hold 25 or 50 stocks.
Transaction costs are not included in most tests but are discussed as a real-world factor.

2.4 Statistical Measures

Compound Annual Growth Rate (CAGR): the primary return measure.
Standard Deviation: annualized volatility.
Sharpe Ratio: risk-adjusted return (excess return over risk-free rate / standard deviation).
Maximum Drawdown: worst peak-to-trough decline.
Base Rate: percentage of all rolling periods where the strategy beats the benchmark. A base rate >70% over rolling 5-year periods is considered strong.

3. Value Factors

3.1 Price-to-Sales (P/S) — The King of Value Factors

O'Shaughnessy's most important finding: P/S is the single best value factor for predicting future stock returns.

Performance by P/S decile (All Stocks universe, 1964–2009):

Decile 1 (cheapest P/S):  ~18% CAGR
Decile 10 (most expensive P/S): ~4% CAGR
Spread: ~14% per year

Why P/S works best:

Sales are harder to manipulate than earnings.
P/S is always positive (companies always have sales, even when earnings are negative).
P/S works in cyclical downturns when earnings have temporarily collapsed.
Extremely high P/S ratios (>10x) are almost always a sell signal.

Implementation rule: Buy stocks with P/S in the bottom decile (lowest 10%). Avoid stocks with P/S in the top decile.

3.2 Price-to-Earnings (P/E)

The most well-known value factor. Results:

Low P/E decile: ~16% CAGR
High P/E decile: ~8% CAGR
Spread: ~8% per year

Limitations:

P/E is undefined for loss-making companies (must exclude them).
Earnings are easily manipulated through accounting choices.
One-time items can distort P/E.
Works well but is strictly inferior to P/S as a single factor.

3.3 Price-to-Book (P/B)

Low P/B decile: ~15% CAGR
High P/B decile: ~9% CAGR
Spread: ~6% per year

Limitations:

Book value is increasingly distorted by intangible assets, share buybacks, and goodwill.
Asset-light businesses (software, services) often have meaningless book values.
P/B is the weakest of the major value factors in O'Shaughnessy's data.

3.4 Price-to-Cash-Flow (P/CF)

Low P/CF decile: ~17% CAGR
High P/CF decile: ~6% CAGR
Spread: ~11% per year

Cash flow is harder to manipulate than earnings.
P/CF is nearly as effective as P/S and often used as a confirming factor.

3.5 EV/EBITDA

Low EV/EBITDA decile: ~17.5% CAGR
High EV/EBITDA decile: ~5% CAGR
Spread: ~12.5% per year

Enterprise value accounts for debt, making it superior to price-only metrics for comparing companies with different capital structures.
EBITDA strips out non-cash charges, providing a cleaner earnings measure.
EV/EBITDA rivals P/S as the best single value factor.

3.6 Value Factor Summary Ranking

Rank	Factor	Approximate Spread	Reliability
1	P/S	14%	Very High
2	EV/EBITDA	12.5%	Very High
3	P/CF	11%	High
4	P/E	8%	High
5	P/B	6%	Moderate

4. Growth Factors

4.1 Earnings Growth Rates

O'Shaughnessy's key finding on growth: past earnings growth does NOT predict future stock returns.

Buying stocks with the highest 5-year earnings growth rates produces below-market returns.
The problem: high growth attracts attention, drives up prices, and the growth rate mean-reverts.
Investors systematically overpay for past growth.

Top decile by 5-year EPS growth: ~11% CAGR (underperforms market)
Bottom decile by 5-year EPS growth: ~14% CAGR (outperforms market)

This is one of the most counterintuitive findings in the book.

4.2 Persistent Earnings Growth

One-year earnings growth alone has limited predictive power.
But consistently increasing earnings (earnings up for each of the past 5 years) does add value when combined with other factors.
The key is consistency, not magnitude. A company growing earnings 10% every year for 5 years is more attractive than one growing 50% then -30% then 40%.

4.3 Sales Growth

Sales growth is a somewhat better predictor than earnings growth because sales are harder to manipulate.
But like earnings growth, extremely high sales growth rates mean-revert.
Moderate, consistent sales growth (5-15% per year) combined with cheap valuation is the sweet spot.

4.4 Growth at a Reasonable Price

Growth alone does not work.
Growth combined with reasonable valuation (low P/S, low EV/EBITDA) works very well.
The best growth stocks to own are those that are still cheap — they have not yet been "discovered" by the market.

5. Momentum and Relative Strength

5.1 The Momentum Effect

After value, momentum is the second most powerful stock market factor. O'Shaughnessy measures it as 6-month or 12-month relative price strength — how a stock's price performance ranks relative to all other stocks.

Top decile by 6-month relative strength: ~16% CAGR
Bottom decile by 6-month relative strength: ~7% CAGR
Spread: ~9% per year

5.2 Why Momentum Works

Underreaction: the market is slow to fully incorporate new information. Positive news leads to a series of upward price adjustments rather than an immediate repricing.
Herding: as a stock rises, more investors take notice, creating buying pressure that perpetuates the move.
Confirmation bias: investors who already own the stock seek confirming information and hold.

5.3 The Interaction Between Value and Momentum

This is O'Shaughnessy's most important finding about combining factors:

Cheap stocks with strong momentum produce the best returns.
Cheap stocks with weak momentum are often value traps — cheap for a reason, going lower.
Expensive stocks with strong momentum can work short-term but are high-risk.
Expensive stocks with weak momentum are the worst performers.

                    Strong Momentum    Weak Momentum
Cheap (Value)        BEST (~20%+)      Value Trap (~12%)
Expensive (Growth)   Risky (~14%)      WORST (~4%)

5.4 Relative Strength Implementation

FUNCTION relative_strength_rank(stock, universe, lookback=6_months):
    all_returns = [s.price_return(lookback) FOR s IN universe]
    stock_return = stock.price_return(lookback)
    percentile = rank(stock_return, all_returns) / len(all_returns)
    RETURN percentile  # 1.0 = best performer, 0.0 = worst

Rule: Only buy stocks with relative strength percentile > 0.70 (top 30%).

6. Shareholder Yield

6.1 Beyond Dividend Yield

Traditional dividend yield screens miss a crucial form of capital return: share buybacks. O'Shaughnessy defines shareholder yield as:

Shareholder_Yield = Dividend_Yield + Buyback_Yield

Where:
Buyback_Yield = (Shares_Outstanding_Prior_Year - Shares_Outstanding_Current) / Shares_Outstanding_Prior_Year

6.2 Why Shareholder Yield Is Superior

Many companies prefer buybacks to dividends for tax efficiency.
A company buying back 5% of its shares annually is returning as much capital as a 5% dividend — but it does not show up in dividend yield screens.
Shareholder yield captures both forms of capital return.

Top decile by shareholder yield: ~17% CAGR
Bottom decile (diluters): ~6% CAGR
Spread: ~11% per year

6.3 Net Share Issuance as a Negative Signal

Companies that are net issuers of shares (diluting shareholders) consistently underperform.
This includes companies using stock options heavily, making acquisitions with shares, or doing secondary offerings.
Avoid any company with net share issuance > 5% of outstanding shares.

7. Small Cap vs Large Cap

7.1 The Size Effect

Small Stocks (All Stocks universe): ~13.5% CAGR, 20% standard deviation
Large Stocks: ~11.2% CAGR, 16% standard deviation

Small caps outperform but with significantly higher volatility.
Maximum drawdowns for small caps are also larger (-55% vs -45%).
The small cap premium has diminished in recent decades as the effect became well-known.

7.2 Micro-Caps: Beware

Stocks below $200M market cap show even higher returns in backtests but are largely untradeable.
Illiquidity, wide bid-ask spreads, and market impact costs make micro-cap backtests unreliable.
O'Shaughnessy excludes micro-caps from his universe for this reason.

7.3 The Optimal Approach

Apply factor strategies within both small and large cap universes.
Factor premiums (value, momentum) are larger in small caps, making the combination extremely powerful.
But position sizing and liquidity constraints must be respected.

8. Cornerstone Value Strategy

8.1 Definition

The Cornerstone Value Strategy is O'Shaughnessy's signature value strategy. It selects stocks from the Large Stocks universe using simple, clear criteria:

UNIVERSE: Large Stocks (market cap > market cap average)

FILTERS:
1. Market cap > database average (large cap filter)
2. Number of common shares outstanding > average
3. Cash flow per share > average
4. Sales > 1.5x average
5. Dividend yield > average

FROM qualifying stocks, SELECT top 50 by SHAREHOLDER YIELD (highest)
REBALANCE annually

8.2 Performance

Cornerstone Value CAGR (1964-2009): ~15.2%
Large Stocks benchmark CAGR: ~11.2%
S&P 500 CAGR: ~10.5%

Sharpe Ratio: ~0.62 (vs 0.40 for S&P 500)
Maximum Drawdown: ~-38% (vs -45% for S&P 500)

8.3 Why It Works

The financial strength filters (high cash flow, large company, high sales) eliminate weak companies.
The shareholder yield ranking rewards companies actively returning capital to shareholders.
The combination produces a portfolio of financially strong, undervalued, shareholder-friendly large caps.
The strategy is easy to implement and maintain with annual rebalancing.

8.4 Behavioral Advantage

The strategy is boring. It buys large, well-known companies that are temporarily out of favor.
This boredom is an advantage: few investors get excited about boring value stocks, so the mispricing persists.

9. Cornerstone Growth Strategy

9.1 Definition

The Cornerstone Growth Strategy is O'Shaughnessy's signature growth strategy. It selects stocks from the All Stocks universe using growth AND value criteria:

UNIVERSE: All Stocks (market cap > $200M inflation-adjusted)

FILTERS:
1. Market cap > $200M
2. Earnings per share growth persistent (EPS higher than prior year for multiple years)
3. Price-to-Sales ratio < 1.5

FROM qualifying stocks, SELECT top 50 by 6-MONTH RELATIVE STRENGTH (highest)
REBALANCE annually

9.2 Performance

Cornerstone Growth CAGR (1964-2009): ~16.8%
All Stocks benchmark CAGR: ~13.5%

Sharpe Ratio: ~0.55
Maximum Drawdown: ~-42%

9.3 Why It Works

The P/S < 1.5 filter ensures you are not overpaying for growth. This single filter eliminates the biggest risk in growth investing.
The earnings persistence filter identifies companies with genuine, sustained growth — not one-hit wonders.
The momentum overlay (relative strength ranking) acts as a timing mechanism, ensuring you buy growth stocks that the market is beginning to recognize.
The combination avoids the classic growth trap: paying 10x sales for a high-growth company that disappoints.

10. Multi-Factor Models

10.1 The Power of Combining Factors

O'Shaughnessy's most important practical finding: combining multiple factors produces better risk-adjusted returns than any single factor alone.

Single factor returns are good. Two-factor returns are better. Three or more factors approach the optimal frontier.

10.2 Two-Factor Combinations

Best two-factor combinations (from All Stocks universe):

Combination	Approx CAGR	Sharpe
Low P/S + High Relative Strength	~19%	0.68
Low EV/EBITDA + High Relative Strength	~18.5%	0.65
High Shareholder Yield + High Relative Strength	~18%	0.63
Low P/CF + High Relative Strength	~17.5%	0.60

The pattern is clear: value + momentum is the winning combination.

10.3 Compositing Value Factors

Rather than relying on a single value metric, O'Shaughnessy recommends a value composite that averages rankings across multiple factors:

FUNCTION value_composite_rank(stock, universe):
    ps_rank = percentile_rank(stock.p_s, universe, ascending=True)  # lower = better
    pe_rank = percentile_rank(stock.p_e, universe, ascending=True)
    pcf_rank = percentile_rank(stock.p_cf, universe, ascending=True)
    ev_ebitda_rank = percentile_rank(stock.ev_ebitda, universe, ascending=True)
    pb_rank = percentile_rank(stock.p_b, universe, ascending=True)
    sy_rank = percentile_rank(stock.shareholder_yield, universe, ascending=False)  # higher = better

    composite = AVERAGE(ps_rank, pe_rank, pcf_rank, ev_ebitda_rank, pb_rank, sy_rank)
    RETURN composite

The composite value rank produces more consistent results than any single value metric because it diversifies across the weaknesses of individual metrics.

10.4 Three-Factor Model

FUNCTION three_factor_score(stock, universe):
    value_score = value_composite_rank(stock, universe)       # Lower = cheaper
    momentum_score = relative_strength_rank(stock, universe)  # Higher = stronger
    quality_score = shareholder_yield_rank(stock, universe)   # Higher = better

    # Combine (normalize each to 0-100 scale)
    combined = (100 - value_score) + momentum_score + quality_score
    RETURN combined

11. United Cornerstone Strategy

11.1 Definition

The United Cornerstone Strategy combines Cornerstone Value and Cornerstone Growth into a single portfolio:

Portfolio = 50% Cornerstone Value + 50% Cornerstone Growth
REBALANCE annually

11.2 Performance

United Cornerstone CAGR (1964-2009): ~16.0%
Sharpe Ratio: ~0.64
Maximum Drawdown: ~-35%

11.3 Why Combining Works

Value and growth (with momentum) have different return drivers and different periods of outperformance.
When value struggles (late-stage bull markets), growth-momentum tends to do well.
When growth-momentum struggles (sharp reversals), value tends to hold up.
The combination smooths the equity curve and improves the Sharpe ratio above either component alone.

11.4 Practical Benefits

Lower tracking error vs. the market (easier for investors to stick with).
Lower maximum drawdown than either strategy alone.
More consistent year-to-year returns.
Reduced behavioral risk — investors are less likely to abandon the strategy during inevitable rough patches.

12. Sector Analysis

12.1 Sector Concentration Risk

Factor-based strategies can produce extreme sector concentrations. A low P/E screen in 2008 would have been heavily weighted toward financials — a disaster.
O'Shaughnessy recommends monitoring sector weights but not imposing hard constraints unless concentrations are extreme.

12.2 Sector-Relative Value

An alternative approach: rank stocks on value within their sector rather than across the market.

FUNCTION sector_relative_value(stock, sector_peers):
    # Rank stock's P/S within its own sector
    sector_ps_rank = percentile_rank(stock.p_s, sector_peers, ascending=True)
    RETURN sector_ps_rank

This prevents the portfolio from becoming a bet on one or two sectors.

12.3 Which Sectors Value Works Best In

Value factors are most effective in cyclical sectors: industrials, energy, materials, consumer discretionary.
Value factors are less reliable in financial sectors (where book value can be misleading) and technology (where traditional metrics may not capture growth optionality).
Momentum factors work across all sectors but are particularly strong in technology and healthcare.

13. Risk and Drawdown Analysis

13.1 Maximum Drawdowns by Strategy

Strategy	Max Drawdown	Recovery Time
S&P 500	-45%	~5 years
All Stocks (value)	-50%	~4 years
Large Stocks (value)	-38%	~3 years
Cornerstone Value	-38%	~3 years
Cornerstone Growth	-42%	~3.5 years
United Cornerstone	-35%	~3 years

13.2 Base Rates

Base rate = percentage of rolling N-year periods where the strategy beats the benchmark.

Strategy                  1-Year   3-Year   5-Year   10-Year
Cornerstone Value         62%      72%      82%      91%
Cornerstone Growth        64%      74%      85%      93%
United Cornerstone        66%      78%      88%      96%

The longer the holding period, the more certain the outperformance. This is why patience and discipline are paramount.

13.3 Worst Periods

Every strategy has multi-year periods of underperformance. Even the best strategies can lag for 3-5 years.
O'Shaughnessy stresses: the ability to stick with a strategy through its worst periods is the primary determinant of investment success.
Investors who abandon value during late-1990s tech bubbles or growth-momentum during sharp reversals destroy their long-term returns.

"The biggest mistake investors make is not understanding that strategies that work over the long term can, and do, have horrible short-term returns."

14. What Doesn't Work

14.1 Factors with No Predictive Power

O'Shaughnessy tested many popular strategies that turned out to have no reliable edge:

Low P/E alone (without other quality filters): works but is weaker than P/S.
High earnings growth without valuation constraint: fails. High-growth stocks are overpriced and underperform.
Low price (penny stocks): no return premium; higher risk.
Analyst recommendations: no predictive value. By the time a consensus buy is issued, the move has happened.
Year-over-year earnings surprises (without other factors): weak and inconsistent.

14.2 Factors That Hurt Returns

High P/S ratios (>10x): consistently terrible. Among the strongest negative signals in the data.
Net share issuers: companies diluting shareholders underperform by 3-5% annually.
Low or negative relative strength: stocks with the worst momentum continue to underperform for 6-12 months.
High debt-to-equity: highly leveraged companies have wider drawdowns and lower risk-adjusted returns.

14.3 The Biggest Myth

"The biggest myth in investing is that you should buy great companies. The data clearly shows that buying good companies at great prices vastly outperforms buying great companies at good prices."

16. Key Principles Summary

P/S is the king of value factors. If you use only one valuation metric, use price-to-sales. It is the most reliable predictor of future returns across all time periods and market environments.
Value + Momentum is the optimal combination. Cheap stocks with strong relative price strength produce the best risk-adjusted returns. Neither factor alone is as powerful as the combination.
Past earnings growth does not predict future returns. Investors systematically overpay for past growth. Buy moderate, consistent growth at cheap prices instead.
Shareholder yield beats dividend yield. Include buybacks in your yield calculation. Companies reducing share count are creating value; companies diluting are destroying it.
Composites beat single factors. Average multiple value metrics (P/S, P/E, P/CF, EV/EBITDA, P/B, shareholder yield) into a composite rank for more consistent results.
The United Cornerstone Strategy is the practical answer. 50% Cornerstone Value + 50% Cornerstone Growth, rebalanced annually. Simple, effective, and robust.
Discipline is the real edge. Every strategy has multi-year periods of underperformance. The investor who sticks with a proven quantitative strategy through the bad times will outperform the investor who chases last year's winner.
Avoid stocks with high P/S ratios. Stocks trading above 10x sales are among the worst investments in the entire database. This is the single strongest negative signal.
Base rates matter more than individual outcomes. A strategy that wins 85% of rolling 5-year periods is excellent. You do not need to win every year — you need to win most multi-year periods.
Simplicity wins. The best strategies in the book are also among the simplest. Complexity does not add returns — it adds fragility and opportunities for error.

"Indexing works because most investors cannot bring themselves to systematically follow any strategy, no matter how well it has performed in the past. The emotions of investing are the investor's worst enemy."