Indicators and Their Limits

The short answer

An indicator takes price (and sometimes volume), applies arithmetic, and plots the result. Whatever comes out was already in what went in.

This is not a criticism. Re-presenting data is genuinely useful — a number between 0 and 100 is easier to compare across instruments than a raw price series. But it sets a hard ceiling on what any indicator can do, and most misuse comes from acting as though the ceiling is not there.

Two families

Trend-following indicators — moving averages, MACD, ADX — smooth price to describe direction and strength. They lag by construction (see moving averages) and they perform badly in ranges.

Oscillators — RSI, stochastics, CCI — express where price sits within a recent range, usually on a bounded scale. They are more informative in ranges and misleading in trends, because a strong trend pins them at an extreme for as long as the trend lasts.

Notice the symmetry: each family fails precisely where the other is useful. This is why people run both, and why running both produces contradictory readings a good deal of the time. There is no configuration that resolves this. The contradiction is real information about conditions — it usually means the market is neither cleanly trending nor cleanly ranging.

"Overbought" is a description of the formula

RSI above 70 means price has closed up more, and by more, than it has closed down over the lookback window. That is the entire content of the reading.

It does not mean the asset is expensive. It does not mean buyers are exhausted. It does not mean a reversal is due. In a strong trend, an oscillator can remain "overbought" for weeks, and selling into that reading is one of the more reliably expensive habits in retail trading.

The word is a label attached to a threshold someone chose. Treat it as a description of the calculation, not a verdict on the market.

Why stacking does not confirm

If RSI, stochastics and MACD all suggest the same thing, that feels like three witnesses agreeing. It is not. All three are computed from the same closing prices. Their agreement is arithmetic, not evidence.

Genuine confirmation requires sources that could disagree — for example, price structure and something not derived from price, such as volume or a measure of positioning. Even then, "confirmation" is a weaker concept than it sounds, because you choose which sources to consult and when to stop consulting them.

The honest version of this practice is: state in advance which two or three inputs you will look at, and commit to acting on them whether they agree or not. What makes stacking a trap is not the number of indicators; it is adding one more until the chart says what you already believed.

Divergence

Divergence describes a case where price makes a higher high while an indicator makes a lower high (or vice versa). It is a real and observable relationship.

Its relationship to outcomes is much weaker than its popularity suggests. Divergence appears frequently in trends that continue for a long time afterwards, and the same setup will be labelled "hidden divergence" or "failed divergence" depending on what happened next. Any concept with a post-hoc label for every outcome is unfalsifiable, and unfalsifiable concepts cannot be tested — which means they cannot be trusted either.

If you use divergence, define in advance what would make it wrong. If you cannot, you are not using a tool; you are using a vocabulary for narrating whatever occurs.

A workable position

• Use few indicators. Two is usually more than enough; the marginal one mostly adds confidence rather than information.
• Know the formula of every indicator on your chart. If you cannot state what it computes, you cannot know what its readings mean.
• Use indicators to describe conditions, not to generate signals.
• Test any rule you build out of sample, with the method in the backtesting lesson.

The most common upgrade in a trader's chart setup is subtraction.