Martingale on a 2× S&P fund, starting with $10,000
Double your money in every time the position falls another 5% below your average entry.
1. The money doubles. The price moves don't.
Each rung costs twice the last one — but buys you a smaller and smaller slice of price.
Rung 1 fires at −5%. Rung 5 needs −43.5%. Rung 9 needs −82.8%. The gaps widen because averaging down drags your entry price with you — so you pay exponentially more for linearly spaced damage. That is the entire trap in one line.
2. What real history actually asked for
SSO, 2016–2026. Shaded bands are rungs; height is cash deployed.
597 rounds over 10.8 years. Most closed at break-even within days, which is exactly why it feels like it works. The deepest was rung 5 in March 2020 — $320,000 committed against a $10,000 plan.
3. By scenario
2018 and 2022 were survivable if you had 32–64× your stake lying around. COVID needed 256×. A 2008-scale decline needs 512×.
Why "just cap it" doesn't fix this. Capping the ladder converts the martingale into ordinary averaging-down, which this project tested separately: the deep-dip version lost, and the shallow version made +0.8%. The unbounded tail is the only thing that makes the win rate look high — remove it and you remove the strategy.
And the win rate is high. 597 of 597 rounds here closed at break-even, because the sample contains no decline deep enough to break it. That is survivorship in a single instrument: the strategy has a 100% hit rate right up until it doesn't.
Table view
Prices: Alpaca SIP + IEX, split and dividend adjusted, 2016-01-04 – 2026-11-13. The scenario rows past 2022 are computed from the drawdown identity, not observed — SSO launched 2006 but this data begins 2016.