Nine indicators equals zero indicators

If you need six confirmations before entering, you don't have six edges. You have one idea and five ways to arrive late. Most indicator stacks are the same input transformed six times. They agree because they're all reading the same price series, so "confluence" is mostly autocorrelation wearing a costume. I'm not saying use none. I'm saying be able to state, in one sentence, what each one contributes that the others don't. If you can't do that for all six, you have fewer than six.
I
IndicatorSkeptic
11 posts · 0 followers
+ Follow

Related reading

3 replies

DataDrivenDee· Aug 2026 ago
Testable version: remove one and see if the results change. Most people have never run that experiment.
F
FiveMinFiona· Aug 2026 ago
I went from five to two and my results didn't change, which was humbling and also freeing.
BacktestBetty· Aug 2026 ago
Ablation is the right test but do it properly or you'll fool yourself: 1) drop each one individually and record the change, 2) then drop pairs, because two weak ones can be carrying each other, 3) run it on a slice of data you didn't use to pick the stack in the first place. The other check I'd add is just correlating the signal outputs rather than the prices - if two of them fire within a bar or two of each other 90% of the time, that's your answer without any equity curve. And the trap is that adding a sixth almost always improves the backtest slightly, which is exactly what overfitting looks like from the inside.
Sign in to reply →
← All strategies & backtesting