MACD vs RSI
MACD subtracts one moving average from another and has no upper or lower limit, while the relative strength index compresses recent gains against recent losses onto a nought-to-one-hundred scale. Both read the same closing prices, so neither confirms the other.
The two most heavily covered indicators anywhere. They are usually run together and described as confirming each other, which is the one thing they cannot do.
What each one is
MACD is a difference between two moving averages, plotted with a signal line in its own panel. It has no upper or lower limit, so its scale depends on the instrument. MACD covers it.
The relative strength index compresses recent gains against recent losses onto a scale from nought to one hundred, so a reading is comparable across instruments. The relative strength index covers the calculation.
Both take closing prices as their only input. Nothing else enters either calculation, which is the fact that governs everything below.
Where they differ
Whether the scale is bounded. One can print any value; the other cannot leave its range. That makes a reading on the bounded one comparable between instruments and the other one not.
What each is actually measuring. MACD measures whether two averages are separating, which is closer to acceleration. The strength index measures how one-sided recent bars have been.
How each fails in a trend. MACD can stay above its signal line for a long stretch and simply keep going. The bounded one pins at an extreme and stays there, which reads as an urgent signal every bar.
What each is read for. Crossovers on one, threshold levels on the other, and divergence on both — which is where most of the trouble starts.
Where they agree
They read the same input. Both are functions of closing prices, so agreement between them is a property of the arithmetic rather than a fact about the market.
Both lag. Neither can turn before price does, and both are late at every real turn, which is the cost of the smoothing that makes either readable.
Both fail in a range. On this site’s shared series direction runs average 2.01 bars with a longest of 11, and short runs produce repeated false signals in either tool.
And neither contains a stop. The ninetieth percentile bar range here is 1.101, and the invalidation belongs at structure rather than at a line or a threshold.
Which one to use
Run MACD when you want to see acceleration. Whether the gap between two averages is widening or narrowing is a real question, and the bounded oscillator does not answer it.
Run the relative strength index when you want comparability. A reading of eighty means the same shape of thing on any instrument, which the unbounded one cannot offer.
Run one of them, not both. They share an input and a lag, so the second adds a confirmation that was never independent — and the false confidence is worse than having no second tool.
And when you cannot say which question you are asking, run neither yet. Both are answers to a specific question about recent bars, and neither is useful until you know which question that is.
Why they cannot confirm each other
Because independence is what makes confirmation mean anything. Two measurements of the same column of numbers, smoothed slightly differently, will point the same way almost always.
And because the agreement arrives late. Both confirm hardest after a move has run, which is exactly when the distance left to travel is smallest.
What to do with the one you keep
Write the signal down before you look for it. Which cross, which threshold, how many bars — a rule you can state is a rule you can review, and one you recognise on sight is one you will recognise whenever the outcome was good.
Fix the lengths and leave them. Changing them after a losing run produces a fresh set of signals and destroys any ability to judge the old set, so the record never accumulates.
Pair it with something that is not price. If you want an actual second opinion, it has to come from a different input — volume, or the structure of the chart — because anything computed from closes will agree with what you already have.
And keep the count. How many signals fired, how many were taken, what happened. On this site’s shared series 95% of bars sat below a prior peak, so a run of losing signals is the ordinary case rather than evidence the tool has stopped working.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 5 compare the two directly in the
title, at a median of 69 views. Separately, the relative strength index appears in 820 titles at a
median of 5,021 across 529 channels, and MACD in 473 at a median of 3,534 across 339. The counts come
from site/rank_compare.py and site/corpus_count.py.
820 and 473 videos on the two subjects, and 5 comparing them at a median of 69. Roughly 1,300 uploads on the pair and almost nothing putting them side by side — the two most-taught indicators in the corpus, and the question of whether they are independent is essentially unasked.
The answer to the question on that chart is that it is one signal. Both turned because the same closes turned — so the position should be sized on one observation, and the second tool has told you nothing you did not already have.
When it fails
The failure is running both as confirmation, and it produces oversized positions late in moves. The oscillator leaves its threshold, MACD crosses its signal line, and the two are read as separate agreement. They are not separate. Both events happened because the same recent closes moved, so the confidence is doubled while the evidence is not. The trade is entered larger and later than a single reading would have justified, on the strength of an agreement the arithmetic made inevitable.
The second failure is reading the bounded one as overbought. It pins in trends.
A third is trading every crossover. Ranges produce them constantly.
A fourth is comparing MACD levels between instruments. The scale is unbounded.
A fifth is finding divergence by looking. Any two series diverge somewhere.
And a sixth is optimising both. You are tuning one input twice.
Related
MACD covers the unbounded difference of averages. The relative strength index covers the bounded oscillator. And MACD divergence covers the reading both are most often used for.
This pairing is on almost every beginner chart, and the reason given is always confirmation. It is not confirmation. Two calculations run over the same column of closing prices will agree most of the time, and the agreement is a property of the arithmetic rather than evidence about the market.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.