MACD

The difference of two smoothers

RSI measured momentum on a bounded 0–100 scale. MACD measures something different: the spread between two moving averages, smoothed once more for a trigger line, with the spread's rate of change visualized as a histogram. No fixed bound — MACD can go as positive or negative as the difference between its EMAs gets.

Gerald Appel designed it in the late 1970s, and both Murphy and Kaufman treat him as the undisputed originator.

The Moving Average Convergence/Divergence indicator, or simply MACD, was developed by Gerald Appel. — John Murphy

The name is literal: when the two underlying EMAs converge (spread narrowing), the MACD line moves toward zero. When they diverge (spread widening), MACD moves away from zero. That's all it's tracking.

The formula

Three components, one formula stack:

• MACD line — the difference between a 12-period and a 26-period EMA of close. Fast smoother minus slow smoother.
• Signal line — a 9-period EMA of the MACD line. Its own smoother.
• Histogram — the spread between the two lines, drawn as vertical bars above/below zero.

Smoothing type: standard EMAs ($\\alpha = 2/(N+1)$) throughout. Unlike RSI — which uses Wilder's $\\alpha = 1/N$ — MACD uses the plain-vanilla textbook EMA. Both Murphy and Kaufman are explicit about this.

The 12/26/9 numbers are Appel's own defaults and have become an industry default to the point that most charting tools don't let you change them without digging into settings. Appel himself preferred 19/39 when analyzing NASDAQ — a useful reminder that even the inventor didn't treat the defaults as sacred.

Why three lines instead of one?

Because each of the three answers a different question.

And layered under all three: MACD crossing zero is just a mathematical restatement of fast EMA crossing slow EMA — the classic MA-crossover signal from the previous lesson, repackaged. MACD is, at its core, a refined MA crossover system with two extra layers of smoothing bolted on top.

Play with it

Push the market to up. Watch:

• MACD climbs above zero (fast EMA is now above slow EMA).
• Histogram goes positive (MACD above Signal).
• Signal-line crossovers mostly work — each cross up catches the continuation, each cross down catches a pullback.

Now flip to sideways. The same system turns on itself:

• MACD wiggles around zero.
• Histogram chops positive/negative every few bars.
• Every crossover is a losing trade in disguise.

Kaufman's understated framing:

Unfortunately, there were many other crossings that generated losses; therefore, it is necessary to select which trades to enter. — Trading Systems and Methods

Same story as MA crossovers and signal-line crossovers everywhere: no oscillator is a regime filter. You have to know whether the market is trending before the indicator becomes useful.

The four uses of MACD

1. Signal-line crossover (the classic)

The rule every textbook repeats:

Buy when the MACD line (faster) crosses upward through the signal line (slower). Sell when the MACD line crosses below the signal line. — Kaufman

This is the crossover you'll see on every TradingView script, every stock blog, every YouTube tutorial. It's fine. It's also the single noisiest MACD signal — more false positives than true ones in chop.

2. Zero-line crossover

When MACD itself crosses zero. Mechanically equivalent to the fast EMA crossing the slow EMA on the price chart — the thing you spent the last lesson learning to be skeptical of.

Murphy is worth reading twice on this one:

The best buy signals occur below the zero line. The best sell signals come from above. — Murphy

Translation: a signal-line cross that happens in the direction of the zero-line structure is much higher-quality than one that fights it. Buy signals below zero (MACD was negative, now crossing up) in an otherwise stable market are exhaustion/reversal plays; buy signals above zero are continuation plays. Know which you're taking.

3. Histogram turns (the leading edge)

The most underused feature of MACD. Murphy:

Turns in the histogram back toward the zero line always precede the actual crossover signals. Histogram turns are best used for spotting early exit signals from existing positions.

In one of Murphy's case studies: the histogram turned down 10 weeks before the signal-line crossover that finally fired the sell. Ten weeks is a lot of give-back. But the warning is crucial:

It's much more dangerous to use the histogram turns as an excuse to initiate new positions against the prevailing trend.

Translation: histogram turns are early exits, not early entries. They warn that momentum is fading; that's a reason to lighten up on longs, not a reason to short.

4. Divergence

Same idea as RSI divergence, measured on the MACD line (or histogram) instead of RSI.

Hit the button and compare the two price peaks to the two MACD peaks. Price prints a higher high; MACD prints a lower high. The difference between the two EMAs is shrinking even as price keeps grinding up — momentum is leaking out of the trend before price admits it.

Kaufman's framing:

An equally important use of MACD is for divergence signals. Bearish divergence occurs when prices are rising but the MACD values are falling.

His warning about timing, which applies equally to RSI and MACD divergence:

Waiting until the divergence is extremely clear is often too late. Momentum will have achieved most of its correction.

And the bigger warning, about the three sequential signals — divergence, then histogram turn, then signal-line crossover:

By the third sell signal, prices will have already dropped significantly, and you will be disappointed with your entry price and the lack of profit opportunity.

The sequence is: divergence (weeks early, highest risk/reward) → histogram turn (days early, decent risk) → crossover (right at the reversal, lowest edge). Beginners use only the crossover. Professionals read the earlier signals and scale out across them.

The timeframe rule

Same as every other lesson. Quoting Murphy verbatim:

As with all technical indicators, signals on weekly charts are always more important than those on daily charts… the weekly signals become trend filters for daily signals. Two crossover systems in which this principle is especially true are MACD and Stochastics.

Practical workflow:

1. Read the weekly MACD. Is it above or below zero? That's your bias.
2. Take only daily signals agreeing with the weekly bias.
3. If weekly MACD is above zero and rising → only take daily bullish crossovers. Ignore the bearish ones.

This single filter eliminates most of the whipsaw losses a naive daily MACD produces.

The Bulkowski epilogue

Thomas Bulkowski, whose statistical testing of chart patterns we leaned on heavily in the S/R lesson, has a candid note in Encyclopedia of Chart Patterns:

I threw in some indicators … (MACD, Bollinger bands, Fibonacci…), all of which I no longer use. — Bulkowski

He has no dedicated MACD chapter and no statistical tables on MACD signals. His pattern-testing framework repeatedly warns against mixing MACD into pattern-recognition work. That's one serious voice in this field saying that, after running 150,000 samples, he personally doesn't trade off the indicator.

Worth sitting with. It doesn't mean MACD is useless — Murphy and Kaufman would argue otherwise and have their own decades of evidence — but it's a useful counterweight to the cultural worship of the 12/26/9 signal line.

Hidden traps

• Unlike RSI, MACD is unbounded. There's no "overbought at X" reading. Don't invent one.
• Zero is the midpoint, not 50. The analogy to RSI's 50 line is MACD's zero line — same regime-filter logic.
• The defaults 12/26/9 are daily defaults. On intraday timeframes they're often too slow; on monthly they're too fast. Period-match your timeframe.
• MACD is an MA crossover in a trench coat. When you read MACD you're reading a smoothed MA crossover. It has the same regime-dependence problem; do not expect it to magically escape what you already learned.
• Bulkowski's reservation: even a serious backtester who believes in pattern testing doesn't personally use MACD anymore. One signal worth weighing.

Quick check

What you now know

• MACD = EMA(12) − EMA(26). The difference of two trend-smoothers. Signal is a 9-EMA of that difference. Histogram visualizes the spread.
• Standard EMAs throughout — α = 2/(N+1). Not Wilder's smoothing. Don't confuse the families.
• Unbounded. Unlike RSI (0–100), MACD can go as positive or negative as the EMA spread allows. No "overbought at X" reading.
• Zero-line crossovers = MA crossovers. Identical signal, repackaged. Same regime-dependence problem.
• Four uses, ordered from earliest/riskiest to latest/safest: divergence → histogram turn → signal-line cross → zero-line cross.
• Weekly MACD filters daily MACD. Non-negotiable per Murphy.
• Histogram leads crossover. Often by weeks. Use it for early exits, not for fighting the trend with new positions.
• One serious backtester (Bulkowski) has publicly stopped using it. Weigh that against Murphy and Kaufman's long-running endorsement. The answer is probably "useful but not sufficient" — like every other indicator.

Next: Dow Theory — step back from indicators and into the grandfather of all this: Charles Dow's six tenets, published in Wall Street Journal editorials around 1900 and still the most compact statement of how trends actually work.