How to Choose Which Indicators to Use
Choosing indicators is a question of correlation rather than quality. Five common momentum oscillators measured on the same history agree with each other at r = +0.75 to +0.93, so stacking them adds confidence and no information, while a volatility measure correlates near zero and genuinely adds one.
The usual advice is a list of favourites. The useful question is not which indicator is best but which two are not measuring the same thing, and that has a number.
How it works
Correlation between two indicators, measured across the same 576-bar history, answers the question directly. A pair near +1 rises and falls together, so the second one is telling you what the first already said. A pair near zero is measuring different things.
Five common momentum oscillators, every pair:
| Stochastic %K | CCI 20 | Williams %R | Money Flow Index | |
|---|---|---|---|---|
| RSI 14 | +0.87 | +0.91 | +0.88 | +0.76 |
| Stochastic %K | +0.89 | +0.93 | +0.82 | |
| CCI 20 | +0.89 | +0.75 | ||
| Williams %R | +0.76 |
Not one pair below +0.75.
Different formulas, different names, different scales — and the same information. Three of them on a chart is one indicator drawn three ways, and what they add is agreement, which feels like evidence and is not.
What a second opinion looks like
RSI against ATR measures −0.08.
A trend measure against ATR measures 0.00.
That is a genuine second axis. ATR does not know which way price is going and momentum does not know how far it moves per bar, so the two together describe something neither describes alone — which is why the stop-distance arithmetic on this site is always ATR-based and never momentum-based.
The rule that falls out of this is short: pick indicators that disagree with each other by construction. One direction measure, one volatility measure, and stop.
Lookback beats formula
A fast exponential pair and a slow simple pair — different formulas and different lengths — correlate +0.30.
Compare that with the +0.87 between two momentum oscillators of the same length.
The length you choose separates two indicators far more than the formula does. Two tools looking at the last 10 bars will agree whatever their maths; the same tool at 10 bars and at 50 bars will not.
So the second lever is lookback, and it is the one people change least. The multi-timeframe page is the same idea taken to the chart itself.
Where volume sits
On-balance volume against RSI measures +0.55.
Halfway. It carries information price alone does not — it uses the volume column — but it is constructed from the direction of the closes, so it inherits most of what momentum already told you.
Which places the three families in order: a direction measure, then volatility as the clear second, then volume as a partial third. Anything beyond that is duplicating one of them.
Confirmation, measured
The reason people stack indicators is confirmation — only take the signal when a second tool agrees. That is testable too.
Starting from the 26 MACD crosses on the do indicators work page, which alone gave 14 wins and 11 losses:
| Filter added | Signals left | Won | Lost |
|---|---|---|---|
| None | 26 | 14 | 11 |
| Momentum must agree (RSI above 50) | 17 | 8 | 8 |
| Volatility must be above its median | 15 | 7 | 8 |
Both filters removed about a third of the trades. Neither improved what was left.
That is the cost of confirmation stated plainly: you give up a third of your opportunities for a result that, on this data, was no better. It may be worth it on other data — but it is a trade, and it is usually described as if it were free.
The correlations above explain part of it. A filter that agrees with your signal 87% of the time cannot change much; it can only remove the cases where the two disagreed, and there is no reason those were the bad ones.
A worked example
List what is on your chart.
Correlate each pair across a long history. Any spreadsheet will do it; the series are just columns of numbers.
Anything above about +0.7 is a duplicate — keep the one you understand best and delete the other.
Then look for a gap. If everything left measures direction, the missing axis is volatility, and that is the one that sets your stop.
The original data
Across our study of 24,971 trading videos, 32 cover how to choose indicators. The median one gets 5,780 views, and only 59% fail to pass 50,000 — a comparatively strong result on one of the smaller fields measured here.
The corpus carries description text for 16 of those 32, and across those 16, not one mentions invalidation, failure, or what a bad read looks like.
Worth putting beside the field sizes. 32 videos on which indicators to choose, against 1,008 on ICT concepts and 966 on scalping — the question of what belongs on the chart is roughly one thirtieth as popular as the answers.
When it fails
Correlation is not the only reason to keep two
Two tools can agree almost always and disagree at exactly the moment that matters. A +0.9 pair is still a duplicate for practical purposes, but the argument is statistical rather than absolute, and it is stated that way deliberately.
The numbers are from one series
A different instrument gives different correlations. The method transfers; the exact figures above do not, and the point of the worked example is to run it on your own data.
You cut down to one
The argument is against duplicates, not against having two. One direction measure with no volatility measure leaves the stop distance to be guessed, which is the more expensive mistake.
You judged them afterwards
Every indicator can be shown working on a chosen chart. The correlation test is worth running precisely because it does not depend on choosing an example.
Related
Do indicators work is the prior question, and why this page treats them as measurements.
ATR is the second axis — the one measure on this page that is not correlated with the others.
And RSI is the momentum family’s most-used member, explained properly.
I run two things on a chart and that is a deliberate limit rather than a preference. What made me cut it down was noticing that whenever my three momentum tools disagreed I ignored the odd one out, and whenever they agreed I felt certain - so they were never changing a decision, only changing how sure I felt about it.
— Michael Whitman, from this video
This page is educational, not financial advice. Test every idea on your own charts before risking money.