Study one
The Factor Reference
Twelve equity risk premia built from the Fama-French research portfolios, scored on return, risk, statistical strength, cost drag, drawdown, post-publication decay — and whether they survived out-of-sample since 2017.
Factor scorecard
Click any column header to sort. Net return applies a turnover-based cost drag (0% market, 1%/yr low-turnover, 3.5%/yr high-turnover). OOS = annualized since 2017-03.
A t-stat above ~3 is the modern bar for believing a factor is real rather than data-mined (Harvey, Liu & Zhu, 2016). Compare with the folklore battery, where the best external signal never survived multiple-comparison correction.
Growth of $1
Log scale. Click a legend entry to show or hide a series.
Rolling 10-year Sharpe ratio
Every factor spends long stretches at or below zero — the part investors actually live through.
Out-of-sample since 2017-03
Annualized return in the ~9 years after the original frozen dataset ended — the honest test of whether a premium is still paying.
The market premium and small-cap momentum kept paying; the low-volatility proxy and large-cap value did not. Out-of-sample is where data-mined factors go to die.
What happened after publication
Annualized return before each factor's founding paper versus after it.
Annualized return by decade
Green positive, red negative. Hover for detail.
Correlation matrix
Monthly return correlations. Orange positive, blue negative.
Low pairwise correlation — value vs momentum especially — is the strongest practical argument for combining factors.
Maximum drawdown
Peak-to-trough loss and when it happened.
Every factor here has lost more than a third of its value at some point.
Rule-based timing on the S&P 500
Long-only switching rules, net of 5bps per unit of position change. Idle cash earns nothing.
Nothing here makes more money than buy-and-hold; the 200-day average makes less-bad drawdowns.
Valuation: CAPE and a simple timing overlay
Shiller's cyclically-adjusted P/E, and a strategy that scales equity exposure by how cheap CAPE is versus its own history (lagged, no look-ahead).
Expensive markets have delivered lower subsequent returns on average, but CAPE is a weak market-timing tool: the overlay's value is drawdown reduction, not extra return.
Factor returns by market regime
Annualized factor return when the market is 20%+ off its peak (bear) versus otherwise (calm).
The defensive factors earn their keep exactly when the market does not.