How should you evaluate an investment strategy backtest?
A backtest asks how a set of rules would have behaved in historical markets. It can reveal useful patterns, but it is a research result. It is not a live investment record or a promise about the future.
A convincing chart can still be misleading if the strategy accidentally uses information that was only available later. Point-in-time index membership, historical prices, publication timing, and realistic portfolio rules all affect whether the result resembles a decision that could actually have been made.
Use an appropriate benchmark
Compare the strategy with a realistic alternative over the same dates. Starting Index uses SPY as a broad-market reference for its S&P 500 stock universe.
Inspect drawdowns
Total return does not show the full experience. Maximum drawdown helps describe the deepest historical decline and the risk an investor would have needed to tolerate.
Look across years
Annual returns reveal whether a result depends on a small number of unusually strong periods and how the model behaved across different markets.
Separate research from live data
A backtest is calculated backward, while a live record accumulates after publication. Reconstructed profile history belongs in its own clearly labeled category.
- Look-ahead bias: using information before it would have been available.
- Survivorship bias: testing only companies that survived to the end of the period.
- Overfitting: tuning rules until they explain the historical sample unusually well.
- Execution assumptions: ignoring turnover, timing, spreads, taxes, or other implementation friction.
- Selective reporting: showing the best model or date range without the weaker alternatives that were tested.
The Starting Index evidence page includes model and SPY backtest curves, annual returns, CAGR, maximum drawdown, point-in-time constituent coverage, and validation sweeps. The audit page separately documents publication dates, benchmark metadata, artifact freshness, and source hashes.