Trading Indicators

What an indicator is

A chart shows prices. An indicator is a formula applied to those prices, or to volume, and drawn on top of the chart or below it. A moving average is the simplest one: add up the last N closing prices and divide by N. The result is a smoother line than the raw price.

The formula is fixed and the output is exact. What is not fixed is the meaning people give it. An average that slopes up is a fact about the last N bars. Calling that "a trend that will continue" is a belief, and this page sticks to the facts.

The four kinds

Trend

Trend indicators ask which way price has been moving. Moving averages, EMA crossovers, the golden cross, Supertrend and the Ichimoku cloud belong here. They are slow on purpose, because smoothing is what removes the noise.

Momentum

Momentum indicators ask how fast price has been moving and whether that speed is changing. The RSI and the MACD are the usual examples. They tend to turn before price does, but they also give more false turns.

Volatility

Volatility indicators ask how big the moves are, ignoring direction. Bollinger bands and the average true range are the common ones. A bot often uses them for sizing and stop distance rather than for direction.

Volume

Volume indicators ask how much traded. Plain volume, the VWAP, the anchored VWAP and the volume profile fit here. Volume data depends on the market. Some markets report one shared total and some do not, so check what your data source actually covers.

Common indicators at a glance

Each row is a one line meaning of the formula, not a rule for trading.

Lines drawn at levels, such as support and resistance, and the breakout indicator are close relatives. They are built from the same price history and share the same limits.

Why indicators lag

An average of the last N closes cannot move until new closes arrive, and it keeps old closes in the mix until they drop out. So it always trails price. A longer window is smoother and trails more.

The picture shows a made-up step. Price sits at 100 for eight bars, then jumps to 110 on bar 9 and stays there. A 5-bar average reaches 110 only on bar 13, 4 bars after the step. A 10-bar average reaches it on bar 18, 9 bars after. No setting is both smooth and instant. Less lag means more noise, and less noise means more lag.

Why stacking indicators adds little

A common idea is that two or three indicators agreeing is stronger than one. That only holds if they carry different information. Most do not. A 5-bar average, a 10-bar average, an RSI and a MACD are all built from the same closing prices. When price rises, they all rise.

Take a made-up path of 24 closes that drifts up from 100 to 121. Over the last 15 bars, the 5-bar and 10-bar averages have a correlation of 1.00. Each one also has a correlation with the price itself of at least 0.97. A second average mostly repeats the first. A bot that waits for both to agree is mostly asking the same question twice.

Mixing kinds helps a little more, such as a trend line with a volume measure, because they draw on different data. But each added condition is another setting, and each setting is another chance to fit the past. See overfitting.

How a bot turns an indicator into a rule

A bot cannot read a chart. It needs a condition that is true or false on each bar. The usual forms are a comparison, such as price above the average, a crossing, such as the fast line passing the slow line, or a level, such as RSI below a number.

Here is a made-up rule on the step path: be long when the close is above the 5-bar average, and flat otherwise. On bar 9 the close is 110 and the average is 102.0, so the rule says long. On bar 10 the close is 110 and the average is 104.0, still long. The rule stays long until the average catches up, then the close equals the average and the rule is no longer long. In a gradual move it would enter later, and in a sideways market it would flip back and forth.

How bots break on indicators

• Fitting the settings. Trying many lengths until one looks good on past data finds noise, not a pattern.
• Using a signal in the wrong market. A rule built for ranges fails in trends, and the reverse.
• Ignoring costs. A rule that looks fine before costs can lose after them.
• Using a bar before it closes. An indicator on an unfinished bar changes until the bar ends. A test that uses the final value early cheats. See look-ahead bias.

What indicators do not tell you

They do not say what price does next. They do not account for news, costs or your account size. They describe the past and give a bot a clear, testable condition. That second use is real. The first use is where beliefs creep in.