“I’ll just combine a few signals”
The intuition feels ironclad: if one indicator is noisy, require three to agree and the noise cancels. We tested this exhaustively — consensus voting, indicator-plus-filter, two-indicator combos, across timeframes and 391 coins (OKX USDT-SWAP universe, June 2026).
What we found
- 5-signal consensus: 55% win rate — and still a loss after stripping market beta.
- Bollinger + Ichimoku, 6 variants: 0 of 6 positive. Adding the Ichimoku trend filter performed identically to ignoring it.
- 4 indicators × 3 filters × 3 timeframes (36 cells): 0 of 36 positive. The “+α” confirmation filters did not improve the base at all.
The reason: correlated signals stack into beta
Here is the math that the intuition misses. RSI, MACD, moving averages, Ichimoku — they are all derived from price. They are highly correlated with each other. When you combine correlated signals, you do not cancel noise; you amplify the common component, which is exposure to the market itself (beta).
Diversification only reduces risk when the things you combine are uncorrelated. Stacking five price-derived indicators is like asking five people who read the same newspaper for independent opinions. You get one opinion, louder — not five independent ones.
We measured this directly: a basket of directional strategies had an “effective rank” far below its count, and every component, after a beta-strip, was the same disguised market exposure.
When combining actually helps
Combination creates value only across uncorrelated return streams. The one pairing that worked in our tests was volatility-selling (VRP) with a trend overlay — genuinely different sources of return, unlike price-derived indicators stacked on each other. That is diversification. Stacking five momentum-flavored indicators is not. (A correlation figure used to sit here; it is gone. The research ledger public/data/research-ledger.json records vrp-shortvol with a verdict of STRONG_LEAD and the date 2026-06-20 and no correlation field — we cannot say which measurement that number came from. The four other figures in this article are owned by that ledger.)
Test your own stacked strategy on the simulator and check the beta-stripped result, not the raw curve. More indicators that all watch price is more beta, not more edge.
So what does work?
If indicators, patterns, and copy-trades all fail an honest test, the obvious question is: then what? Our answer isn’t a sharper prediction — it’s risk management. After testing 17 strategy families across 15 research campaigns, the only thing that survived wasn’t forecasting the next move; it was controlling how you hold — sidestepping the worst drawdowns and surviving the cycle. That’s crisis defense, not a crystal ball, and we never call it more than it is.
- The honest answer: a risk-managed portfolio — survive the bear, compound through the cycle, half the drawdown of buy-and-hold.
- The proof, in the open: our trust page — every result, the failures included.
- Check it yourself: the simulator — your strategy, real fees, real coins.
Don’t believe us. Verify.