I backtested over a dozen trading ideas. One survived. What idea did real data kill for you?

Every one of them looked good on paper. Scalping on order flow, mean reversion, the Turtle rules across 19 markets, a Sunday gap trade in forex, funding and ML filters on top of a trend strategy.

Real data killed all of them. The one that survived is a plain trend follower on ETH: +79.5% over 6.5 years in backtest. Buy and hold made far more, but its worst drop was 79%. The bot's was 6.9%.

The rule I ended up with: an idea isn't dead until I test it on data it never saw, and it isn't alive until it beats a random baseline. What's an idea you were sure about that data killed? Doesn't have to be trading.