Long-Only Factor Portfolios: Beyond Fama-French

Backtesting 13 non-traditional return drivers as long-only top-10 portfolios in the S&P 500. Low Vol and Asset Turnover stand out on a risk-adjusted basis.

As a follow-up to the factors we explored previously, this post reviews 13 non-traditional return drivers as simple long-only, top-10, monthly-rebalanced portfolios in the S&P 500 — a back-of-envelope exercise to see how different fundamental and technical signals map to return patterns. These are illustrative, not backtests of implementable strategies: no transaction costs, no survivorship correction, no multiple-testing adjustment. Treat the numbers as directional.

The Return Drivers

Return DriverCategoryCore IdeaMeasurementIdeal Values
High EBIT/EVGrowthEarnings before interest and taxes relative to enterprise valueEBIT / enterprise valueHigh
FCF/PriceValueFree cash flow relative to stock priceFCF / stock priceHigh
Cash/AssetsValueProportion of assets held in cashCash / total assetsHigh
Inventory ChangeFundamentalsEfficiency of inventory managementChange in inventory over periodHigh
Asset TurnoverFundamentalsEfficiency of asset utilisationRevenue / average assetsHigh
ROEFundamentalsProfitability relative to shareholder equityNet income / shareholder equityHigh
ROAFundamentalsProfitability relative to total assetsNet income / total assetsHigh
EBITDA MarginFundamentalsOperating profitability excluding certain expensesEBITDA / revenueHigh
ROCEFundamentalsEfficiency of capital employedEBIT / capital employedHigh
MomentumTechnical/PricePrice trend continuationRate of change in price over timeHigh
Low BetaTechnical/PriceLower systematic risk relative to marketBeta estimateLow
Low VolTechnical/PriceLower return volatilityStandard deviation of returnsLow
Price/SalesTechnical/PriceMarket capitalisation to total salesMarket cap / total salesHigh

Methodology

For each return driver, I rank all companies in the S&P 500 from best to worst and go long the top 10, rebalancing monthly. The investment universe is S&P 500 constituents at each point in time who have at least 1 year of price data and who are not in the top 10% of firms ranked by beta — there’s evidence that extremely high-beta firms generally don’t do well in the long run, so I filter those out across all tests. The backtest period is January 2000 to July 2023.

Results

Avg. Ann. ReturnAvg. Ann. StdAnnual SharpeWorst Loss 1YWinning Years
Inventory Change10.30%20.50%0.50-52.11%79.17%
Asset Turnover11.52%14.74%0.78-29.84%87.50%
ROE11.42%17.08%0.67-25.21%79.17%
ROA9.90%19.56%0.51-49.22%83.33%
EBIT/EV9.32%25.21%0.37-52.24%75.00%
EBITDA Margin8.76%23.35%0.38-62.64%70.83%
FCF/Price10.34%29.09%0.36-70.58%62.50%
Low Beta8.87%11.03%0.80-23.18%83.33%
ROCE7.36%17.63%0.42-54.32%83.33%
Momentum7.37%24.29%0.30-77.60%75.00%
Low Vol10.17%11.43%0.89-13.87%75.00%
Cash/Assets8.95%24.97%0.36-73.99%70.83%
Price/Sales10.35%28.25%0.37-71.79%75.00%

What Stands Out

A few observations from the results:

Low Vol dominates on a risk-adjusted basis. A 0.89 Sharpe with a worst 1-year loss of only -13.87% is remarkable for a 10-stock portfolio. Low Beta tells a similar story (0.80 Sharpe, -23.18% worst loss). These two factors deliver their returns through volatility compression rather than raw alpha — useful if your binding constraint is drawdown tolerance.

Asset Turnover is the quiet winner among fundamentals. An 11.52% annualised return with 14.74% volatility and 87.5% winning years is an attractive profile. The -29.84% worst year is manageable. Efficiency of asset utilisation appears to be a meaningful and persistent signal.

Momentum is poor in long-only form — a 0.30 Sharpe and -77.60% worst year. Momentum tends to work better as a long-short factor where the short leg captures the reversal in losers. In long-only form, you’re only getting half the story.

These are 10-position concentrated portfolios, so the results are inherently noisy. Adding more positions would smooth returns but dilute the factor signal. The right portfolio size is a function of your conviction in the signal and your tolerance for tracking error — a tradeoff worth thinking about carefully before implementation.