Investment strategies, backtested against the S&P 500

  • What this is: stock investment strategies, each backtested against the same benchmark.
  • The benchmark: the same amount of money put into the S&P 500 (SPY) and held.
  • The test: each strategy starts as a hypothesis about what to buy, when to buy it and when to sell. It is then replayed over historical prices.
  • A pass: beating SPY on return, on drawdown, or both.
Method
  • No survivorship bias: stock pickers may only buy companies that were in the S&P 500 on that date.
  • No hindsight bias: fixed baskets are tested only from the date they were published.
  • Execution: trades fill at the next day's open, at a cost of 0.1% per trade.
  • Prices: adjusted for splits and dividends.
  • Measures: annual return (CAGR) and maximum drawdown (worst peak-to-trough drop), each next to SPY over the same period.

Findings to date

  • Stock picking: none of the 6 strategies tested beat SPY once bias was removed.
  • Index timing: 2 of the 4 strategies beat SPY over 1994–2026, at 12.0–12.8%/yr vs 10.9%. Their worst drop was −34% vs −55%. Both trailed SPY from 2010 onward.
  • Limits: that edge comes from sidestepping two bear markets (2001–02 and 2008), which is a small sample. Past results do not guarantee future returns.
Trend + Sahm rule Index timing

Best result so far: 4 trades in 32 years, beat SPY with ~40% less drawdown. Edge rests on 2 recessions.

12.8%/yr vs SPY 10.9%

worst drop -33.7% vs -55.2%

▲ Promising
Growth-trend timing (unemployment) Index timing

Beat SPY with ~40% less drawdown by stepping aside in 2001–02 and 2008; slightly behind since 2010.

12.0%/yr vs SPY 10.9%

worst drop -33.7% vs -55.2%

▲ Promising
Trend filter (Faber 10-month) Index timing

Cut the worst drawdown by more than half, but gave up ~1.3%/yr and whipsawed in 2022.

9.6%/yr vs SPY 10.9%

worst drop -24.6% vs -55.2%

◆ Insurance only

Failed

Hypotheses rejected: each underperformed the same investment in SPY once tested without bias.

Graham margin of safety Value investing

Its picks lagged: fully invested it made 8.3%/yr with a −51% drawdown; with T-bill padding (shown) 10.4%/yr. Only 0–11 names pass a year.

10.4%/yr vs SPY 15.1%

worst drop -20.6% vs -33.7%

✕ Avoid
Trend + yield curve Index timing

Stayed invested through 2008 — the curve inverted more than a year earlier.

10.3%/yr vs SPY 10.9%

worst drop -49.5% vs -55.2%

✕ Avoid
CAN SLIM (price-action proxy) Stock picking

Its apparent edge was survivorship bias — bias-free it trailed SPY by ~7%/yr.

6.8%/yr vs SPY 13.9%

worst drop -29.2% vs -33.7%

✕ Avoid
Minervini Stage 2 Stock picking

Trailed SPY by ~8%/yr once survivorship bias was removed, with deeper drawdowns.

5.7%/yr vs SPY 13.9%

worst drop -41.6% vs -33.7%

✕ Avoid
Sin stocks (vice) Fixed basket

Trailed SPY by ~10%/yr with a deeper drawdown; the real vice fund (VICEX) lagged SPY by ~4%/yr over 24 years. Even consumer staples did better.

3.5%/yr vs SPY 13.8%

worst drop -55.1% vs -33.7%

✕ Avoid
Graham defensive investor Value investing

Almost never finds a stock: 0–1 S&P 500 companies passed all seven tests in any year 2012–2026, so it sat in T-bills.

2.6%/yr vs SPY 15.1%

worst drop -4.9% vs -33.7%

✕ Avoid
Becky 10 (lifestyle brands) Fixed basket

Reproduces the famous +1,000% for 2015–20, but from the day the list was fixed it lost money while SPY made 15%/yr — and trailed consumer discretionary (XLY) and QQQ too.

-0.6%/yr vs SPY 15.0%

worst drop -57.6% vs -24.5%

✕ Avoid