MACD — Moving Average Convergence Divergence
The answer
MACD is the gap between a 12-period exponential moving average and a 26-period one. It measures whether the shorter-term average is pulling away from the longer-term average or closing in on it — in plain words, whether a trend is speeding up or slowing down. It is built from moving averages, so it lags, and in a sideways market it crosses back and forth producing losses.
Why this costs you money
MACD is usually taught in the same paragraph as the word "momentum", and the word does the damage. People treat it as a leading indicator. It is not one, and it cannot be one.
Every input to MACD is a moving average of past prices. A moving average cannot turn before price does. Two moving averages subtracted from each other cannot turn before price does either. The histogram is a moving average subtracted from a subtraction of moving averages, and it also cannot.
Here is the specific loss. A stock ranges sideways for 3 months. The 12 and 26 averages sit almost on top of each other, so MACD hovers around 0. Every small move produces a crossing of the signal line. In 3 months of sideways price, a default MACD will typically produce 6 to 10 crossovers. Almost all of them lose, because the moves are small and the entry is late by construction.
The second loss is larger and it happens once. In a fast crash, MACD's sell signal arrives after most of the fall. In a fast recovery, its buy signal arrives after a large part of the rise. A trader who uses MACD crossovers as their primary entry in volatile conditions will be systematically late on both sides, and being late on both sides of a round trip is the definition of losing money in a market that went nowhere.
MACD's honest job is to describe whether a trend is strengthening or weakening. Used as a trigger for entries, it charges you the full lag of a 26-period average every time.
How it works
Three parts. Each measures something different, and most people use only the first 2.
1. The MACD line. Take a 12-period exponential moving average of the close. Take a 26-period one. Subtract the second from the first.
MACD line = EMA(12) − EMA(26)
When the MACD line is positive, the short average is above the long one, which means recent prices are above older prices. When it is rising, the gap is widening, which means the trend is accelerating.
2. The signal line. A 9-period exponential moving average of the MACD line. It is slower. It exists to give you a crossing point.
3. The histogram. The MACD line minus the signal line, drawn as bars.
Histogram = MACD line − signal line
The histogram is the most interesting of the 3 and the least used properly. It does not measure the trend. It measures the rate of change of the gap. When the bars are getting taller, the trend is accelerating. When they are shrinking, the trend is still going in the same direction but is decelerating.
What each reading actually says
| What you see | What it literally means |
|---|---|
| MACD line above 0 | The 12 EMA is above the 26 EMA. Recent prices exceed older prices. |
| MACD line crosses 0 | The 12 EMA has crossed the 26 EMA. A moving average crossover, renamed. |
| MACD crosses above signal | The gap between the averages has started widening upward |
| Histogram shrinking while positive | Still an uptrend, but decelerating |
| MACD far from 0 | The 2 averages are unusually far apart |
Notice the second row. A MACD zero-line crossing is exactly the same event as a 12/26 moving average crossover. It is not additional information. It is the same information plotted differently, which is easier to see and no more predictive.
Why 12, 26 and 9
Gerald Appel developed MACD in the late 1970s. The 12, 26 and 9 settings are widely reported to come from a 6-day trading week, when Saturday sessions existed: 12 was 2 weeks, 26 was roughly a month, 9 was a week and a half. If that account is right, the global default setting on every charting platform is calibrated to a trading calendar that no longer exists anywhere.
Whether or not the story is true, the point stands. The numbers are conventions. Change them and the signal count changes.
What it tells you, and what it does not
MACD tells you the distance between 2 moving averages and whether that distance is growing or shrinking. Everything else is inference.
It does not lead. Ever. It is arithmetic on past prices.
It is unbounded. Unlike RSI, MACD has no ceiling. This has 1 advantage and 1 serious limitation. The advantage is that it does not saturate — it can keep showing more strength. The limitation is that there is no level that means "extreme". A MACD of 40 is meaningless without knowing what MACD values that instrument normally reaches, and MACD values scale with the price of the instrument. A 3,000-rupee stock and a 30-rupee stock cannot have their MACD values compared at all.
It has no meaning in a range. When price oscillates around a level, the 2 averages converge on that level, MACD sits near 0, and the signal line crosses constantly. Every crossing looks like the crossings that occur in trends.
Its divergence is a warning, not a trigger. MACD divergence — price making a higher high while MACD makes a lower high — says the second advance was made with the averages closer together than the first. In a mature trend that is normal, because the largest percentage moves usually come early. Divergence appears many times in a strong trend, and most instances mean nothing. It tells you where to look. Price structure tells you when.
The clean statement of the limit: MACD is a description of trend health. Any use of it as a trigger inherits the lag of a 26-period average, and you should know what that lag costs before you accept it.
The decision rule
Use the histogram to describe the trend, and use the zero line as a filter. Do not use the signal-line crossover as your entry.
If the MACD line is above 0, take long setups from your price structure and skip short ones. If it is below 0, reverse it.
If the histogram is shrinking while the MACD line stays above 0, the trend is decelerating. That is a reason to tighten a stop or to stop adding, not a reason to reverse.
Unless the MACD line has been within a narrow band around 0 for more than a month. That means the 2 averages are sitting on top of each other and price is ranging. In that condition MACD has no content, and every crossover it produces should be ignored.
Try this now
This takes 5 minutes and it measures the lag in a number rather than in a word.
- Open a 1 year daily chart of a stock from your own watchlist that had at least 1 clear trending move. Add MACD (12, 26, 9) with the histogram shown.
- Find the 3 largest price swings on the chart — the 3 clearest moves from a turning point to a turning point.
- For each swing, mark 2 dates: the date price actually turned, and the date the MACD line crossed the signal line in the new direction.
- Count the number of trading days between the 2 dates. Then work out how much of the swing had already happened by the crossover date, as a percentage.
- Write down 3 numbers. For example: 6 days late, 22% of the move gone. 9 days late, 31% gone. 4 days late, 15% gone.
What you should see. The crossover will always arrive after the turn — never before, on any chart, in any market. That is not bad luck, it is the construction. The interesting number is the percentage. On most charts you will find that between 15% and 35% of each swing was already complete before MACD confirmed it.
That percentage is the price of the tool, stated in the only unit that matters. If a typical swing on your instrument is 12% and MACD costs you 25% of it, MACD costs you 3 percentage points of every trade. Now you can decide whether the confirmation is worth 3 points. Before this exercise, you could not.
Bonus, 60 seconds. Change 12 and 26 to 5 and 13. Repeat step 3 on 1 swing. The lag shrinks. Now count the crossovers over the full year at both settings. The faster setting will give you far more of them. You have just bought earlier entries with more false signals, at an exchange rate you can now see.
Three real cases
1. Gerald Appel, late 1970s (United States) — the settings nobody has retested Gerald Appel, a New York money manager and newsletter writer, developed MACD in the late 1970s. The default 12, 26, 9 has been the shipped setting on charting platforms ever since, applied to every instrument and every timeframe from 1-minute charts to monthly ones. Appel himself wrote about varying the settings for different purposes and used different values for buy and sell signals in some of his work. The developer varied the settings. The users almost never do.
2. The S&P 500, February to July 2020 (United States) — late on both sides of a round trip The index fell roughly a third between late February and 23 March 2020, then recovered to a new high by August. A default daily MACD crossed below its signal line after the decline was already under way, and crossed back above after a substantial part of the recovery had happened. Both crossings were correct. Both were late. A trader taking only those 2 signals sold well below the high and bought well above the low, which on a round trip that ended higher than it started produced a loss on a market that gained.
3. The NIFTY 50, 2022 (India) — a year of crossings and no trend The Indian index moved in a wide choppy range through 2022, falling in the first half and recovering in the second, and ended the year not far from where it started. In that condition the 12 and 26 averages spend most of their time close together, MACD hovers near 0, and signal-line crossings occur repeatedly. Each crossing was arithmetically correct and none of them led anywhere, because there was no trend for a trend tool to measure. The indicator did not fail. It reported, accurately, that the 2 averages kept swapping places, which is what happens when price goes nowhere.
The question that resolves it
A novice sees the MACD line cross above the signal line and asks: is momentum turning up?
An expert asks: where is the MACD line relative to 0, and how long has it been there?
A signal-line crossing that happens well above 0 is a trend that paused and resumed. A signal-line crossing that happens with MACD sitting near 0, for the fourth time in 3 months, is 2 moving averages sitting on top of each other in a range. The crossing looks the same on the chart. It is not the same event.
What would make this wrong
The claim is that MACD lags by a measurable amount, that it has no content when price is ranging, and that its zero-line crossings are simply moving average crossovers renamed.
You would falsify the lag claim by finding a chart where the MACD crossover preceded the price turn. It cannot happen with the standard formula, because every input is a past price, and confirming that on your own chart is worth more than accepting it here.
You would falsify the range claim with the counting exercise. Count crossovers on a ranging chart and on a trending chart over the same year. If the ranging chart did not produce more crossovers, or if those crossovers were followed by moves as often as the trending ones, this article is wrong about your instrument.
The honest limits are 4.
First, this article treats the signal-line crossover as the standard use. Many experienced traders use only the histogram, or only divergence, and those uses are less exposed to the lag.
Second, the lag measurement in the exercise depends on how you identify a "turn", which is a judgement made after the fact. Different people will mark the turns differently and get somewhat different numbers. The order of magnitude is the useful part, not the exact figure.
Third, being late is not automatically bad. Late entries have a higher hit rate than early ones, because some of the uncertainty has resolved. The exercise measures the cost. It does not measure what you bought with it.
Fourth, nothing here shows that MACD adds anything over a plain 12/26 moving average pair on the price chart. On the zero-line crossing, it demonstrably does not — it is the identical event. The signal line and histogram are the parts that carry something extra, and whether that something is worth a panel of screen space is a fair question for any reader to answer with their own count.
In India
Indian platforms ship MACD at 12, 26, 9 and Indian traders use it heavily on daily and weekly charts, and on 15-minute charts for intraday work.
Three Indian conditions matter.
MACD values are not comparable across stocks, and this bites harder in India. MACD is expressed in the price units of the instrument. The Indian market has a very wide spread of share prices, from single-digit rupee stocks to shares costing tens of thousands of rupees. A MACD reading of 40 is enormous on one and invisible on the other. Any screener that ranks Indian stocks by MACD value is ranking them by price level, not by momentum. If you want to compare, you need a percentage version of MACD, and most platforms do not offer one by default.
Price bands stretch a move across several days. A share outside the derivatives segment carries a daily band of 2%, 5%, 10% or 20%. A move that wanted to happen in 1 session arrives across 3. Moving averages actually handle this reasonably well — the staircase feeds smoothly into the EMAs. The problem is that the MACD confirmation arrives even later than usual, because the move itself is being delivered slowly. In a circuit-locked stock you cannot act on the confirmation anyway, since there is no liquidity at the band.
Weekly MACD is popular in Indian technical commentary. A weekly 12, 26, 9 crossing on the NIFTY 50 is reported as a major event in Indian financial media. It is worth knowing what that actually is: a 26-week exponential average and a 12-week one, so roughly a 6-month and a 3-month view, confirming something that started around 2 months earlier. It is not news. It is a slow summary of old news, and it is treated as a headline.
In the United States
Same defaults, same 3 parts, and 2 differences.
Results-day gaps enter the averages whole. Most large US companies report after the close, and shares trade in extended hours that the standard daily candle excludes. A stock that moves 15% overnight delivers that entire move to the EMAs as 1 close-to-close change. The 12-period average, which weights recent data heavily, jumps. MACD widens sharply in a single day. In India that same move would arrive across several banded sessions and the widening would be gradual.
No daily price bands means larger MACD excursions. With single-session moves uncapped, US MACD values swing further from 0 than comparable Indian ones do on similar news. Any threshold you have read in an American book about "MACD extremes" is calibrated to that.
The US also gives you the cleanest place to run the lag exercise at scale. If you want to know the average lag of a 12/26/9 crossover across 300 liquid stocks over 10 years, US data is where you can compute it, mostly for free.
Where they differ, and what that tells you
The difference is how fast the input arrives.
In the United States, a large repricing lands in 1 close-to-close change, so the exponential averages absorb it immediately and MACD reacts within a day or 2.
In India, price bands deliver the same repricing across several sessions, so the averages absorb it gradually and MACD reacts over a week.
What that tells you is that MACD's lag is not a fixed number of days. It is a function of how the market delivers information. A trader who measures MACD's lag on US charts and applies the figure to Indian ones will be consistently optimistic about how quickly the confirmation arrives.
It also tells you something about which market MACD suits better. MACD's real strength is describing whether a sustained trend is accelerating or decelerating. Indian banded delivery actually produces smoother trends in banded stocks, which suits the tool. But the same bands mean you cannot trade the confirmation, since a stock at its band has no liquidity. So MACD's description is more accurate in India and less actionable, which is an uncomfortable combination and worth knowing before you build anything on it.
Carry this
- MACD is the distance between a 12 and a 26 period average. The zero-line crossing is a moving average crossover with a different name.
- The histogram measures acceleration, not direction. Shrinking bars in an uptrend mean the trend is slowing, not reversing.
- Measure the lag on your own chart as a percentage of the swing. That number is what MACD costs you, and you should not use it without knowing it.