← sources

Survivorship bias can really wreck a backtest

source: reddit (reddit:algotrading)published 2026-10-10 13:48ingested 2026-10-11 02:42

retained text

I ran a little experiment using a Dow 30 rotation strategy. First I backtested it using today's Dow 30 stocks, then ran the exact same strategy using the actual historical Dow components at each point in time. The results were dramatically different. The first backtest looked great, but the historically accurate version was pretty dismal. The animation shows the difference. It's a good reminder of how misleading a backtest can be when you're using today's index constituents to test strategies going back years.

Curious how others handle this who backtest using stock index components. Do you use historical constituent data, or just accept survivorship bias as a limitation?

document fetched 2026-10-11 00:40 · canonical https://reddit.com/r/algotrading/comments/1x2frbt/survivorship_bias_can_really_wreck_a_backtest