What an indicator is, and what it cannot be

An indicator is a calculation. It takes numbers your chart already has — open, high, low, close, volume, sometimes time — and returns a different number. That is the whole of it. Nothing enters an indicator that was not already on your screen.

This sounds obvious written down, and almost nobody trades as though they believe it. The moment a line appears in its own panel, in its own colour, with its own scale, it starts to feel like a separate source of information — a second opinion. It is not a second opinion. It is the same data, rearranged.

Three things a calculation can usefully do

Nearly every indicator you will meet is doing one of three jobs. Knowing which one you are looking at is more useful than knowing its name, because it tells you immediately what to expect and what to stop expecting.

Smoothing

Removing detail so a shape becomes visible. Moving averages do this, and so does almost every line drawn over price. The cost is always the same and it is not negotiable: what you gain in clarity you pay for in delay.

Normalising

Rescaling something so it can be compared with itself at another time, or with another instrument. Oscillators do this — they take a quantity with no natural bounds and express it between 0 and 100. The cost is that the scale is relative to a window you chose, so «high» means high compared with the recent past, and nothing more.

Accumulating

Adding up something that happens too fast or too small to watch, so the total becomes legible. Volume-based measures do this. The cost is that a running total carries its history forever, so where you started it changes what it says.

A large part of the disappointment people feel with indicators comes from expecting the wrong job. Asking a smoother to be early is asking it to stop smoothing. Asking a normaliser to tell you direction is asking a thermometer for the weather forecast.

Lag is arithmetic, not a defect

This is the single most useful thing on this page, so it gets its own section.

The average of the last twenty prices is, by construction, describing a point roughly ten bars ago. Not because the formula is crude, but because that is what an average of twenty things is.

Every «improvement» on the moving average — weighting the recent values more heavily, adapting the period, filtering differently — is trading one part of that cost for another. Faster response means more of the noise gets through. Smoother output means more delay. You can choose your position on that line. You cannot leave it.

When you read that some indicator is «leading» rather than «lagging», the honest translation is almost always: it responds to smaller changes, so it moves earlier, and it is wrong more often. That can be a good trade. It is still a trade.

Where the settings came from

The default numbers on your indicators are not discoveries. 14, 20, 12 and 26 are conventions, most of them chosen decades ago by people working with daily bars, hand calculations or early computers, for markets that no longer behave the way those markets behaved.

That does not make them wrong. Widely used settings have one genuine advantage that has nothing to do with the arithmetic: other people are looking at the same lines, and in a market of participants that is not nothing. But it does mean that hunting for the «correct» period is hunting for something that does not exist. What exists is a setting suited to a chart, a session and a question — and if you cannot say which question, the number is decoration.

There is a particular trap here worth naming, because it costs real money and looks like research: testing many settings on past data and keeping the one that would have worked best. On any data set, something always would have worked best. Finding it tells you about that data set.

What no indicator does

Being specific about this is more respectful than being vague, so:

• It does not know what has not traded. Resting orders, size waiting at a price, someone about to change their mind — none of that is in the open, high, low and close. That information exists, but it lives in the order book and the tape, which is a different part of this path.
• It does not know why. A calculation cannot distinguish a move caused by a scheduled release from an identical move caused by one large participant leaving.
• It does not predict. Every value it has ever printed was computed from bars that had already closed. A line that turns up did so because the recent past turned up.
• It does not know your account. The same reading is a reasonable risk for one person and an unreasonable one for another, and the indicator has no way to tell.

None of this makes indicators useless. It makes them instruments. A well-built instrument shows you something you would otherwise have to hold in your head, faster and with less effort — and holding less in your head is the entire point, because attention is the resource you actually run out of during a session.

How to read the seven pages after this one

What follows groups them by family: trend, momentum, volatility, then VWAP on its own because its one peculiarity deserves the room, then volume and open interest, then the filters of an engineer who approached the whole problem differently, and finally the retracement grid, which is a different kind of object altogether. Each page says how the thing is calculated, what it is genuinely good at, and — the part most writing leaves out — what it does not tell you.

One of them breaks the rule set out at the top of this page, and it is flagged where it happens: open interest is not computed from your chart. The exchange counts it and publishes it. It is the only figure in the whole section that widens the box on the left of diagram 1, which is precisely why it is worth knowing.

They are not recommendations, and there is no ranking. If you finish them able to say what each one is measuring and what it is blind to, they have done their job.