Value Composite 1, 2 & 3The original O'Shaughnessy strategy. Retested with institutional data.
O'Shaughnessy's What Works on Wall Street reported 18–21% CAGR for the Value Composite. We retested three versions on 21 years of out-of-sample data (2005–2026), using quarterly trailing-twelve-month fundamentals from Sharadar, with no survivorship bias. The results are significantly weaker than the book claims — but there is still alpha, if you know where to look.
What Is the Value Composite?
The Value Composite ranks every stock in the market by multiple valuation metrics simultaneously. The idea is simple: no single metric tells the full story. P/E can be distorted by one-time charges. P/B misses intangible assets. P/S is more stable but ignores profitability. By combining several metrics into a single composite score, the strategy captures "cheapness" across multiple dimensions.
Each stock receives a percentile rank (0 = cheapest, 100 = most expensive) for each metric. The composite score is the average of all percentile ranks. The 25 stocks with the lowest composite — consistently cheap across all dimensions — form the portfolio.
What the book promised: O'Shaughnessy backtested the VC2 on US equities from 1927 to 2009 and reported approximately 18–21% CAGR — roughly double the S&P 500. These results were in-sample (the strategy was developed using that data).
Results — Annual Rebalance
Each strategy rebalanced once per year. To eliminate timing bias, we ran 12 portfolios — one starting in each calendar month — and averaged the results. This ensures no single rebalance month drives the outcome.
>$200M
| VC1 | VC2 | VC3 | S&P | |
|---|---|---|---|---|
| Avg CAGR | +4.9% | +5.6% | +5.2% | ~8.4% |
None of the three composites beat the S&P 500 at >$200M with annual rebalance. The book’s original specification fails out-of-sample.
>$5B
| VC1 | VC2 | VC3 | S&P | |
|---|---|---|---|---|
| Avg CAGR | +9.7% | +9.8% | +9.3% | ~8.4% |
At >$5B, VC1 and VC2 generate modest alpha (+1.3–1.4%). Better, but still far from the book's claims.
VC2 >$200M — by Rebalance Month
| Month | CAGR |
|---|---|
| Jan | +6.6% |
| Feb | +5.7% |
| Mar | +3.6% |
| Apr | +2.8% |
| May | +4.8% |
| Jun | +4.1% |
| Jul | +4.5% |
| Aug | +6.9% |
| Sep | +6.5% |
| Oct | +5.5% |
| Nov | +8.8% |
| Dec | +7.5% |
| Average | +5.6% |
VC2 >$5B — by Rebalance Month
| Month | CAGR |
|---|---|
| Jan | +10.7% |
| Feb | +11.2% |
| Mar | +10.8% |
| Apr | +11.1% |
| May | +9.2% |
| Jun | +9.7% |
| Jul | +9.2% |
| Aug | +9.1% |
| Sep | +8.0% |
| Oct | +8.8% |
| Nov | +9.9% |
| Dec | +10.0% |
| Average | +9.8% |
Results — Monthly Rebalance
Instead of rebalancing once a year, we rebalance every month using the most recent quarterly fundamentals available. This keeps the data fresh — a stock that reported strong earnings last quarter enters the portfolio immediately, rather than waiting up to 12 months.
>$200M
| Strategy | CAGR | Alpha | Sharpe | Sortino | Max Drawdown | $100 → |
|---|---|---|---|---|---|---|
| VC1 (5 factors) | +10.8% | +2.4% | 0.50 | 0.68 | -75.8% | $878 |
| VC2 (+ SH yield) | +12.1% | +3.7% | 0.56 | 0.80 | -66.5% | $1,118 |
| VC3 (+ div yield) | +8.9% | +0.5% | 0.46 | 0.62 | -73.9% | $607 |
| S&P 500 | +8.4% | — | 0.61 | — | — | $510 |
>$5B
| Strategy | CAGR | Alpha | Sharpe | Sortino | Max Drawdown | $100 → |
|---|---|---|---|---|---|---|
| VC1 (5 factors) | +10.2% | +1.8% | 0.56 | 0.67 | -64.4% | $784 |
| VC2 (+ SH yield) | +10.0% | +1.5% | 0.56 | 0.67 | -62.6% | $746 |
| VC3 (+ div yield) | +11.8% | +3.4% | 0.64 | 0.81 | -54.3% | $1,054 |
| S&P 500 | +8.4% | — | 0.61 | — | — | $510 |
Monthly rebalance is dramatically better than annual, especially at >$200M. VC2 goes from +5.6% (annual avg) to +12.1% (monthly). The reason: quarterly fundamentals change every 3 months. With annual rebalance, you hold stocks for up to 12 months with stale data. Monthly rebalance picks up changes within weeks of the quarterly report.
At >$200M, VC2 (shareholder yield) wins. At >$5B, VC3 (dividend yield) wins. The shareholder yield signal is noisier in large caps because buyback programs are lumpy and irregular. Dividend yield is more stable.
Year-by-Year Returns
VC2 >$5B — Monthly Rebalance
| Year | VC2 | S&P 500 | Alpha |
|---|---|---|---|
| 2005 | +36.7% | +8.4% | +28.3% |
| 2006 | +16.9% | +12.4% | +4.6% |
| 2007 | +2.6% | -4.2% | +6.8% |
| 2008 | -51.2% | -40.1% | -11.1% |
| 2009 | +46.3% | +30.0% | +16.3% |
| 2010 | +25.8% | +19.8% | +6.1% |
| 2011 | +9.7% | +2.0% | +7.7% |
| 2012 | +18.1% | +14.1% | +3.9% |
| 2013 | +43.0% | +19.0% | +24.0% |
| 2014 | +22.2% | +11.9% | +10.3% |
| 2015 | -14.3% | -2.7% | -11.5% |
| 2016 | +16.2% | +17.5% | -1.3% |
| 2017 | +23.4% | +23.9% | -0.5% |
| 2018 | +0.3% | -4.2% | +4.5% |
| 2019 | -8.1% | +19.3% | -27.4% |
| 2020 | -4.1% | +15.2% | -19.3% |
| 2021 | +23.3% | +21.6% | +1.7% |
| 2022 | +7.0% | -9.7% | +16.8% |
| 2023 | +17.8% | +18.9% | -1.1% |
| 2024 | +8.4% | +24.7% | -16.3% |
| 2025 | +16.5% | +14.9% | +1.6% |
| 2026 | +5.1% | -5.9% | +11.0% |
VC3 >$5B — Monthly Rebalance
| Year | VC3 | S&P 500 | Alpha |
|---|---|---|---|
| 2005 | +30.8% | +8.4% | +22.5% |
| 2006 | +23.1% | +12.4% | +10.8% |
| 2007 | +9.6% | -4.2% | +13.8% |
| 2008 | -43.7% | -40.1% | -3.6% |
| 2009 | +52.6% | +30.0% | +22.5% |
| 2010 | +21.3% | +19.8% | +1.5% |
| 2011 | +11.5% | +2.0% | +9.5% |
| 2012 | +27.6% | +14.1% | +13.5% |
| 2013 | +31.9% | +19.0% | +12.9% |
| 2014 | +18.1% | +11.9% | +6.2% |
| 2015 | -15.4% | -2.7% | -12.7% |
| 2016 | +23.5% | +17.5% | +6.0% |
| 2017 | +22.0% | +23.9% | -1.9% |
| 2018 | +6.0% | -4.2% | +10.2% |
| 2019 | -10.9% | +19.3% | -30.2% |
| 2020 | +1.7% | +15.2% | -13.5% |
| 2021 | +27.2% | +21.6% | +5.6% |
| 2022 | +14.2% | -9.7% | +23.9% |
| 2023 | +12.0% | +18.9% | -6.9% |
| 2024 | +0.1% | +24.7% | -24.6% |
| 2025 | +17.9% | +14.9% | +3.0% |
| 2026 | +10.0% | -5.9% | +15.9% |
Sector Concentration
Value composites systematically overweight cheap cyclical sectors. Energy (20%), Consumer Cyclical (18%), and Communication Services (16%) dominate the portfolio across all years. Technology is underweighted at 9% — tech companies rarely screen as "cheap" on traditional value metrics because their value comes from intangible assets (IP, network effects, brand) that don't appear on the balance sheet.
This sector tilt is the primary source of both the strategy's alpha and its risk. When energy and cyclicals outperform (2005–2007, 2016, 2022), the VC crushes the S&P. When they crash (2008, 2015, 2020), the portfolio suffers max drawdowns of −54% to −76%. There are no sector caps — the composite buys whatever is cheapest, regardless of concentration.
Sample Holdings — VC2 >$5B (January 2025)
| Ticker | Sector | P/E | P/S | P/B | VC2 Score |
|---|---|---|---|---|---|
| GM | Consumer Cyclical | 4.6 | 0.3 | 0.7 | 2.2 |
| OVV | Energy | 4.9 | 1.0 | 0.9 | 3.8 |
| PVH | Consumer Cyclical | 7.7 | 0.6 | 1.0 | 4.2 |
| PDCE | Energy | 3.4 | 1.5 | 1.4 | 5.4 |
| VLO | Energy | 11.7 | 0.3 | 1.7 | 5.8 |
| IMO | Energy | 7.7 | 0.7 | 1.6 | 6.0 |
| PAA | Energy | 11.6 | 0.3 | 1.2 | 6.2 |
| CHRD | Energy | 8.6 | 1.7 | 0.9 | 6.6 |
| UCM | Utilities | 9.6 | 1.0 | 1.7 | 6.7 |
| MUR | Energy | 10.8 | 1.6 | 1.0 | 6.7 |
| LEA | Consumer Cyclical | 11.6 | 0.3 | 1.3 | 6.8 |
| BG | Consumer Defensive | 11.9 | 0.3 | 1.4 | 7.4 |
| TAP | Consumer Defensive | 12.0 | 1.0 | 0.8 | 7.6 |
| ET | Energy | 10.8 | 0.7 | 1.6 | 7.8 |
| ADM | Consumer Defensive | 15.9 | 0.3 | 1.3 | 8.1 |
| APA | Energy | 4.1 | 1.0 | 1.8 | 8.3 |
| NXST | Communication Serv | 9.0 | 1.0 | 2.4 | 8.5 |
| NOV | Energy | 5.9 | 0.7 | 1.0 | 8.9 |
| WCOEQ | Communication Serv | 18.9 | 0.6 | 0.3 | 9.1 |
| T1 | Communication Serv | 9.8 | 0.6 | 2.0 | 9.4 |
| VIAB | Communication Serv | 6.3 | 0.8 | 1.1 | 9.4 |
| WCC | Industrials | 11.9 | 0.4 | 1.6 | 10.0 |
| SNX | Technology | 15.2 | 0.2 | 1.3 | 10.0 |
| ADT | Industrials | 7.4 | 1.4 | 1.7 | 10.2 |
| HPE | Technology | 9.8 | 0.8 | 1.0 | 10.3 |
Why the Book's Numbers Don't Replicate
1. In-sample vs out-of-sample. O'Shaughnessy developed the VC2 using 1927–2009 data. Our test period (2005–2026) is entirely out of sample. Strategies routinely lose 30–50% of their backtested edge when applied to new data.
2. The value premium has weakened. From the 1930s through the 2000s, cheap stocks consistently outperformed. Since 2010, growth has dominated. The Fama-French HML (value factor) has been flat or negative in many recent years. This isn't noise — it reflects a structural shift toward intangible-asset businesses where traditional value metrics are less meaningful.
3. Data freshness matters. The book likely used annual data with annual rebalance. Our results show that quarterly TTM fundamentals with monthly rebalance significantly improve performance — suggesting that some of the "value premium" comes simply from acting on fresher information.
4. Survivorship bias in historical data. Older datasets have known survivorship issues — bankrupt value stocks are often missing. Our Sharadar dataset includes 12,151 delisted stocks specifically to avoid this bias.