MACD: Two Averages, Three Lines, and One Documented Failure Mode

Three Lines From Two Averages

Fidelity describes MACD as a momentum oscillator primarily used to trade trends, appearing as two oscillating lines without upper or lower bounds. That unbounded quality separates it immediately from RSI, which is confined between zero and 100.

Gerald Appel developed the 12-26-9 settings in the late 1970s, primarily for daily charts. Every default you see in a charting package traces to that choice.

Why the fast average leads

A 12-period average responds to new prices faster than a 26-period one simply because each new price is a larger fraction of the total. When a stock starts rising, the fast average turns up first and the two separate. When the rise stalls, the fast average flattens while the slow one is still catching up, and the gap narrows before price itself turns.

That narrowing is what the histogram shows, and it is the earliest thing MACD offers.

Two Different Crossovers, Two Different Meanings

People say "MACD crossover" as though it names one event. It names two, and they are not equivalent.

The signal line crossover

When the MACD line crosses above its signal line, Fidelity describes the reading as bullish, and crossing below as bearish. This is the frequent one, occurring whenever short-term momentum shifts relative to its own recent average.

The zero line crossover

MACD crossing above zero means the 12-period EMA has crossed above the 26-period EMA — a slower, more consequential event. Fidelity notes that a bullish signal-line crossing is stronger the further below the zero line it occurs, which is a way of saying that a momentum shift starting from a deeply negative position has more room ahead of it.

Distinguishing the two matters because signal-line crossovers are common and zero-line crossovers are not. Treating them as one signal produces far more trades than the underlying trend changes justify.

The Failure Mode Is Named in the Documentation

Most indicator criticism comes from skeptics. This one comes from the broker publishing it.

Fidelity states that during trading ranges the MACD will whipsaw, with the fast line crossing back and forth across the signal line, and that users generally avoid trading in this situation.

That is a substantial admission. A trend-following construction produces its worst output when there is no trend — and a sideways market is not rare. The indicator gives no warning that it has entered this state; the crossings look identical to the ones that precede real moves.

What whipsaw costs

Each false crossing that is acted on incurs a round trip: the spread on entry, the spread on exit, and any commission. A month of choppy trading can generate several such crossings, none of which produce a move large enough to cover their own costs.

This is the practical mechanism by which a strategy that looks reasonable on a chart loses money in an account.

Reading the Histogram Rather Than the Lines

The histogram is the least discussed component and often the most informative, because it turns a comparison into a single quantity.

Its bars measure the distance between the MACD line and the signal line. Growing bars mean the gap is widening — momentum accelerating relative to its own recent average. Shrinking bars mean the gap is closing, and a crossover is approaching.

The second and fourth rows are where the histogram earns attention. A shrinking bar while price is still making new highs is the same observation divergence makes, arriving through a different route and often somewhat earlier.

The caution is that shrinking bars precede crossovers that never complete. Momentum can slow, stabilize and re-accelerate without the lines ever touching, which produces a warning that resolves into nothing.

Divergence, and Why It Ranks Above Crossovers

Fidelity notes that divergence between the MACD and price action is a stronger signal when it confirms the crossover signals.

The construction is the same as with other momentum measures: price reaches a higher high while the indicator does not, suggesting the force behind the advance has weakened. What distinguishes MACD divergence is that the underlying quantity is explicitly the relationship between two trend measures, so a divergence is describing a trend losing its own internal consistency.

The qualification in that sentence is doing real work. Divergence confirming a crossover is treated as stronger than either alone — which is guidance to require agreement rather than to act on the first thing that appears.

Changing the Settings Changes the Instrument

The 12-26-9 defaults are Appel's choices for daily charts, not constants of nature.

Shortening the periods makes every line more responsive and multiplies the crossings. Lengthening them reduces false signals and delays every real one. As with any smoothing tradeoff, there is no configuration that is both fast and reliable, because responsiveness and noise are the same property viewed from two sides.

Fitting parameters until they perform well on historical data is the specific trap. A setting selected because it worked over the last three years is a description of those three years, and the more combinations tried, the more certain it becomes that the winner was chosen by chance.

What the Indicator Structurally Cannot See

• Volume. MACD is computed from price alone, so an advance on heavy participation and one on almost none look identical to it.
• Why the price moved. Earnings, a regulatory decision, an index rebalance and random drift all enter the calculation the same way.
• Gaps. A stock that opens sharply away from the previous close produces a jump the averages absorb gradually, delaying the reading exactly when events move fastest.
• Liquidity. In securities with thin trading and wide spreads, a handful of transactions can shape the price series the indicator is built from.
• Rule-imposed limits. Limit up-limit down bands constrain what prices may print, so the input is shaped by market structure as well as by demand.

The Costs That Decide Whether Any of This Works

An indicator generating frequent signals must clear a cost hurdle, and the regulator requires that hurdle to be spelled out.

FINRA Rule 2270 obliges firms to hand active traders a standardized risk statement before the account opens. Its most quantitative element is a worked illustration in which commissions alone consume $111,360 of annual profit before the trader breaks even.

The rule also requires warning customers against advertising that trumpets outsized gains, and reminding them that their counterparties include professional traders employed by securities firms. A crossover strategy is competing against those people, using an indicator they can also see.

Why a Lagging Indicator Is Not Automatically Useless

The standard objection is that moving averages describe the past, so MACD can only report what has already occurred. That is true and less damning than it sounds.

Every observable market quantity is historical. A price is the record of a transaction that has completed; volume counts trades already made; an earnings figure describes a quarter that has ended. The question is never whether information is backward-looking but whether it is organized usefully.

What MACD organizes is the relationship between two timeframes, which a raw price chart shows only implicitly. Whether the medium-term trend is accelerating or decaying is hard to read off price alone and is exactly what the histogram displays.

The honest limitation is different from the usual one: not that the indicator lags, but that it reduces a company to a price series and then reduces that series to one relationship within it. What it reports is accurate about that relationship, and silent about everything else.

Where It Earns Its Place

Used as a trend description rather than a trading trigger, MACD is informative. The histogram narrowing while price still rises is a real observation about a move losing force. A zero-line crossing marks a meaningful change in the relationship between short and medium-term trend.

Its most defensible use is as one input requiring confirmation from others — the position Fidelity takes on its indicators generally. Its least defensible use is mechanical trading of every signal-line crossing, which is precisely the application the documentation warns produces whipsaw.

The Bottom Line

MACD subtracts a 26-period average from a 12-period one, smooths the result over nine periods, and plots the difference. Everything it reports is derived from prices that have already printed.

Distinguish the two crossover types, since one is common and one is not. Treat divergence as more informative than either. Expect the indicator to perform worst in sideways markets, because its own documentation says so. And run the cost arithmetic before trading any signal, because the spread is charged whether the signal was right or wrong.

For a bounded momentum measure with different failure modes, see RSI; for what each trade costs, see the spread and order book depth.