Correlated indicators and the illusion of confirmation
Three oscillators agreeing is one opinion said three times. Here is how to tell which of your tools are duplicates.
Two indicators are correlated when they tend to move together because they are computed from overlapping inputs in similar ways. Most popular indicators are far more correlated than their different names and shapes suggest.
Families that largely duplicate each other
| Family | Members that heavily overlap |
|---|---|
| Momentum | RSI, stochastics, rate of change, CCI, Williams %R |
| Trend | SMA, EMA, WMA, MACD lines, moving average envelopes |
| Volatility | ATR, standard deviation, Bollinger width, Keltner width |
Within a row, adding a second member buys you very little. RSI and stochastics reach extremes at broadly the same times because both are asking a version of the same question about the same closes.
A test you can run in ten minutes
Put two indicators you use on the same chart and scroll back through a few hundred bars, looking only for disagreement.
- If you find plenty of it, they are reading different things. Keep both.
- If you struggle to find any, one of them is redundant. Remove it and the chart gets more readable at no cost.
This is worth doing with the indicators you have used for years. Familiarity is not independence.
Correlation changes with conditions
Two tools can be nearly independent in quiet markets and lock together in violent ones. That is the worst possible arrangement, because it means your confirmation quietly disappears exactly when the decisions get expensive.
The same effect shows up across instruments. Six positions in six different names feels diversified until a broad move puts every one of them on the same side. Correlation between indicators and correlation between positions are the same problem wearing different clothes.
What actual confirmation looks like
Confirmation requires a source that could have said no. A price-based setup confirmed by unusual volume is confirmation, because the volume reading was free to be ordinary. A price-based setup confirmed by a second price-based tool is not, because that tool was always going to agree.
The habit worth building: before you add anything to a chart, name the circumstance in which it would contradict what you already have. If you cannot name one, do not add it.