Same ten years.
One backtest quietly cheats.
An index gets reviewed twice a year and swaps names out. A name usually leaves after it has already shrunk or de-rated. After it has already hurt.
| Universe the test ran on |
CAGR |
What it actually holds |
| Today's 50 names, run backwards | ~16% | Only the names that survived to today |
| Index as it stood on each date | ~14% | The deletions too, while they were falling |
Roughly 25 large caps cycled through the list over the decade. Drop them and the curve gains about 2 points a year it never earned. The deleted name that halved before it left is the exact loss the survivor list never books.