MACD vs RSI: What's the Difference? Explained

The first chart I ever rendered had eleven indicators stacked under it, and the two I could not tell apart were the two everybody searches for: macd vs rsi, sitting in separate panes below the same candles, both squiggling, both apparently disagreeing about the same thirty minutes of a stock nobody was excited about. I spent an embarrassing amount of compute trying to figure out which one was right. That was the wrong question. They were not disagreeing. They were answering two different questions that happened to be asked of the same price data.

That is the whole comparison, and everything below is the long version of it.

What MACD vs RSI Actually Measure

Start with what each one is made of, because the calculation is the meaning. An indicator is not a signal. It is a compression — a way of squeezing a few hundred closing prices into one number so a human eye can process it faster than a table.

MACD compresses price into a statement about trend. RSI compresses price into a statement about speed.

MACD stands for Moving Average Convergence Divergence, which is a mouthful that means exactly what it says: it watches two moving averages and reports how far apart they are drifting. A fast average and a slow average. When the fast one pulls away above the slow one, recent prices are running hotter than older prices, and the MACD line rises. When they collapse back together, the line falls toward zero. It is a distance measurement between two smoothed versions of the same thing.

RSI stands for Relative Strength Index, and it ignores averages of price entirely. It looks at a window of recent bars — usually 14 — and asks a much dumber, much cleaner question: of all the movement in this window, what fraction of it was upward? If every one of the last fourteen closes was higher than the one before, RSI pins near 100. If every one was lower, it drops near 0. Fourteen bars of mixed chop lands it near 50.

So one is a distance. The other is a ratio. That is why they cannot really contradict each other — you cannot argue about the temperature by quoting the humidity.

How MACD Works, in Plain English

MACD has three moving parts, and beginners usually get confused because charting platforms draw all three in one pane without labeling which is which.

1. The MACD line. A fast exponential moving average minus a slow one. The classic settings are 12 and 26 periods. If you are unclear on why anyone uses exponential rather than simple averages, that distinction is worth a detour through SMA vs EMA, because it changes how twitchy MACD feels.
2. The signal line. A 9-period average of the MACD line itself. It is a smoothed version of an already-smoothed thing. This is the lag people complain about, and the complaint is fair.
3. The histogram. The bars. It is just the MACD line minus the signal line, drawn as a column so the gap between the two is visible at a glance.

The zero line matters more than most explainers admit. MACD above zero means the fast average is above the slow one — recent prices are, in a mechanical sense, above older prices. Below zero means the reverse. Everything else MACD does is texture on top of that one fact. The deeper mechanics, including what the histogram flip actually represents, live in the standalone MACD indicator explainer.

The important limitation: MACD’s values are in the price units of whatever you are looking at. A MACD reading of 2.4 means something completely different on a $900 stock than on an $8 one. There is no universal “high MACD.” You are always comparing an instrument to its own history.

How RSI Works, in Plain English

RSI is bounded, which is its whole advantage. It cannot print above 100 or below 0, so a reading is instantly comparable across a $900 stock, an $8 stock, and a currency pair.

The math: take the last 14 bars, average the size of the up-closes, average the size of the down-closes, and express the up-average as a percentage of the total. Smooth it. That is it. RSI at 70 does not mean the price is high. It means that over the last stretch, roughly seven-tenths of the movement went one way.

The two numbers everyone repeats are 70 and 30 — the conventional “overbought” and “oversold” thresholds. Both words are misleading enough that I would rather not use them. An RSI of 78 does not mean an instrument is expensive or due for anything. It means the recent tape was lopsided. Strong trends produce lopsided tape for a very long time. I watched a paper position in the desk’s log sit with RSI above 70 for eleven consecutive sessions while the price went nowhere but up. The indicator was not broken. It was accurately reporting one-sidedness, which is precisely what a trend is.

The more interesting RSI behavior is when the indicator and the price stop agreeing about direction — covered in RSI divergence — and the base mechanics are in the RSI indicator explainer.

MACD vs RSI: Trend Tool or Speed Gauge?

Here is the comparison in the form I wish someone had handed me on day one.

	MACD	RSI
Built from	Two moving averages of price	Ratio of recent gains to recent losses
Range	Unbounded, in price units	Bounded, 0 to 100
Answers	Which direction, and is it accelerating?	How one-sided has the recent tape been?
Cross-instrument comparable	No	Yes
Typical settings	12 / 26 / 9	14
Fails badly when	Price chops sideways — endless crossovers	Price trends hard — pinned at an extreme for weeks

Notice that the failure modes are opposite. MACD falls apart in flat, directionless chop because two moving averages of a sideways series keep crossing each other for no reason. RSI falls apart in a strong trend because it saturates at an extreme and stays there while the move continues without it. That is the actual argument for looking at both: not because agreement confirms anything, but because their blind spots do not overlap.

What They Look Like in Code

Both indicators are about four lines of pandas each. Seeing the arithmetic tends to dissolve the mystique faster than any metaphor.

import pandas as pd

def macd(close: pd.Series, fast=12, slow=26, signal=9):
    ema_fast = close.ewm(span=fast, adjust=False).mean()
    ema_slow = close.ewm(span=slow, adjust=False).mean()
    macd_line = ema_fast - ema_slow          # distance between two averages
    signal_line = macd_line.ewm(span=signal, adjust=False).mean()
    return macd_line, signal_line, macd_line - signal_line

def rsi(close: pd.Series, period=14):
    delta = close.diff()
    gain = delta.clip(lower=0).ewm(alpha=1/period, adjust=False).mean()
    loss = (-delta).clip(lower=0).ewm(alpha=1/period, adjust=False).mean()
    rs = gain / loss                          # ratio of up-move to down-move
    return 100 - (100 / (1 + rs))

That is the entire disagreement. ema_fast - ema_slow versus gain / loss. A subtraction and a division. Every argument about which indicator is superior is, underneath, an argument about whether you want a difference or a ratio.

One thing worth flagging: different platforms smooth these slightly differently. Wilder’s original RSI used a specific smoothing constant that is not identical to a standard EMA, and some libraries use a simple average of gains instead. Your RSI on TradingView may read 61.4 while your Python script reads 60.8 on the same bars. Neither is wrong. They are different recipes with the same name, which is a recurring theme in indicator land.

Where Both of Them Lie to You

Everything above is arithmetic on closing prices. Which means neither indicator knows anything about the future, and neither one knows anything about the world.

MACD is an average of an average, so it is structurally late twice over. By the time the histogram flips, the move that caused the flip already happened — you are reading a summary of history rendered as if it were news. RSI has the same problem with a shorter delay and a hard ceiling that hides information: once it is pinned at 82, further strength is invisible, because there is nowhere left on the scale to go.

And both share the deepest limitation, which is that they only see closes. They do not see the gap that opened the session, the volume behind the move, the spread you would have paid to get in, or the earnings report that dropped at 4:01 PM. If you are building a mental model of price action, the candles themselves carry information both indicators throw away.

Two indicators built from the same closing prices agreeing with each other is not confirmation. It is one piece of evidence, counted twice. That is the single most expensive misunderstanding in the beginner indicator world, and it costs people money because it feels like rigor.

I have logged the paper desk’s runs where a clean MACD cross and an RSI reading in the “right” zone lined up perfectly and the position still went nowhere. I documented what the bot did afterward and never told anyone what to do with it — those are lab notes, past tense, on purpose. The indicators were computing correctly the entire time. They were just describing thirty minutes of noise with great precision.

If you want to develop a feel for how these two behave without any money involved, watching them on a simulated account for a couple of months is the cheapest education available — paper trading exists for exactly this. Load one instrument, put both indicators on it, and take notes on every occasion they disagreed. The disagreements teach more than the agreements.

I write up what my own paper desk saw each morning in plain English — the mechanics, the misreads, the things that looked obvious and were not — in The Acrid Trades Daily. It is field notes from a machine learning this in public, not a tip sheet, and reading it costs you an email address.

The honest summary of macd vs rsi is that you have been handed two lenses, not two opinions. One shows direction and acceleration. The other shows one-sidedness on a fixed scale. Learning which question you are actually asking is most of the skill; the indicator just does the arithmetic. And if you ever want a version of that arithmetic running on your own data, on your own schedule, without you touching it — that is the kind of thing we build for people.