Relative Strength Index (RSI)
The oversold exit is the strongest daily signal in the whole study. The overbought exit is not its mirror.
+3.0 pts best edge over a matched random entry · 47,705 signals
MACD
The most-tested signal here by sample size. Significant on the long side, and small enough that costs matter more than the signal.
+1.9 pts best edge over a matched random entry · 121,274 signals
Stochastic Oscillator
One side of this indicator has a small edge. The other is the closest thing to pure noise we measured.
+1.1 pts best edge over a matched random entry · 120,676 signals
Stochastic Momentum Index (SMI)
The double-smoothed stochastic. Beats the original on the long side, and unlike the original its short side is not quite dead.
+1.8 pts best edge over a matched random entry · 107,275 signals
Moving Averages
The most famous signal in retail trading, measured properly. It finishes last on daily charts and first on four-hour.
+1.9 pts best edge over a matched random entry · 139,518 signals
Bollinger Bands
Mean reversion at the band is the classic read. Measured, both sides come in under a point.
+1.0 pts best edge over a matched random entry · 91,583 signals
VWAP
Our largest sample by far. The mean-reversion trade works; the reclaim-VWAP trade everyone teaches does not.
+1.6 pts best edge over a matched random entry · 1,091,284 signals
Supertrend
The strongest daily trend signal here, and completely one-sided: the bull flip pays, the bear flip is nothing.
+2.4 pts best edge over a matched random entry · 40,455 signals
Ichimoku Kinko Hyo
Five lines, and only one of them earns its place: the downward cloud break. The famous cross does not.
+1.8 pts best edge over a matched random entry · 149,604 signals
Average True Range (ATR)
The rare signal that is significantly worse than random. ATR is a superb sizing tool and a poor entry trigger.
+1.1 pts best edge over a matched random entry · 67,204 signals
Why a random entry is the only honest benchmark
A win rate on its own means nothing. Indices, gold and crypto rose over 2010–2026, so a random long with a symmetric target won 51.8% of the time while a random short won 46.6%. Any long signal inherits that 5.2-point gap before its indicator does a single thing. Subtract the control column and most indicators contribute between nothing and two points.
That is the number we report on every page here, and it is why several famous signals score worse than their reputation.