How to Use the RSI
To use the relative strength index, read it as a measure of how one-sided recent price movement has been. In a range, readings at the extremes mark exhaustion; in a trend, the same readings mark strength and persist for a long time. The mistake is applying one interpretation to both.
The relative strength index answers one question: how one-sided has recent movement been. Almost every mistake made with it comes from reading it as an answer to a different question.
Before you start
A period you chose deliberately rather than accepted at 14. Shorter reacts faster and produces more extremes; longer smooths and produces fewer. Pick one and be able to say why.
A decision about whether you are trading with the trend or against it. The indicator’s readings invert in meaning between those two, so this has to be settled first.
The instrument’s ordinary bar range, so a signal can be compared to noise. On this site’s shared series the median bar range is 0.493 with a tenth percentile of 0.17 and a ninetieth of 1.101.
The steps
1. Read the number as one-sidedness
A high reading means most recent movement was upward. It says nothing about value, nothing about fairness, and nothing about what happens next.
2. Decide which market you are in before reading a level
In a range, extremes mark turning points. In a trend, extremes mark strength and keep occurring. Applying the range interpretation to a trend is the single most common error.
3. Set the levels to match that decision
In a strong trend, shift the bands — 80 and 40 in an uptrend, 60 and 20 in a downtrend. The defaults were chosen for a general case, not for the market in front of you.
4. Wait for the indicator to leave the extreme, not enter it
An extreme reading can persist for 20 bars or more. Acting on entry into the zone means acting at the start of a run rather than at its end.
5. Use divergence for direction, and nothing else
Price makes a higher high and the indicator does not. That is the one configuration carrying information about direction rather than about the recent past.
6. Confirm against structure before acting
An extreme reading at a level that already mattered is worth something. The same reading in the middle of nowhere is worth considerably less.
7. Place the stop from the chart, not from the indicator
The indicator has no price levels in it. Where the idea is wrong is a structural question, and the answer sets the position size through the usual division.
How to tell it worked
You can state which market type you decided you were in, before you read any level. If you cannot, the level was interpreted after the fact.
Your bands are not 70 and 30 unless you chose them. The defaults surviving a deliberate review is fine; the defaults surviving because nobody looked is not.
A reading that stayed extreme for 15 bars did not produce a trade. That is the discipline working rather than the indicator failing.
And every entry had a stop taken from structure, not from a level on the oscillator.
What it cannot do
It cannot see anything except recent price. No volume, no context, no news. On this site’s shared series a round trip measures about 2% of the median bar range of 0.493, so a strategy taking every extreme reading pays that repeatedly for signals that carry no directional information.
In a quiet market it reaches extremes on almost nothing. A handful of small bars in one direction produces the same reading a violent run would, and the two are not comparable events.
Choosing the period
A shorter period sees fewer bars, so it reaches extremes more often and leaves them faster. A 2-period setting spends most of its life at one end or the other; a 21-period setting reaches an extreme rarely and means more when it does.
The trade is between frequency and information. More signals is not more edge — it is the same edge divided into smaller pieces, each of which still costs a round trip to act on.
Match the period to the holding period. Someone holding for 3 bars and someone holding for 30 are asking different questions about recent movement, and a single default cannot answer both.
Then leave it alone. A period changed after a losing run is fitted to that run, which is the same error as adding filters to a screen until the list looks better.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 154 have an instruction-shaped
title about the relative strength index, at a median of 4,398 views across 128 channels. Divergence
specifically appears in 22 videos at 9,568 and the average true range in 135 at 11,637. The counts
come from site/rank_howto.py, which deduplicates by video id.
154 videos at a 4,398 median — heavy coverage and a low audience per video. That combination describes a crowded topic where most of the coverage repeats the same default interpretation, which is the one this page argues against.
The answer to the question on that chart is that 82 in a trend is a description of strength. On this site’s shared series, direction runs average 2.01 bars and the longest ran 11 — extremes persist, and a reading is not a countdown. The trade being suggested is against the only thing the indicator actually measured, which is that recent movement was one-sided.
When it fails
The failure is a strong trend read with range settings, and it fails repeatedly rather than once. The indicator reaches 75, then 80, then 85, and each reading looks like a better opportunity than the last while price continues in the same direction. Every entry is stopped out and the indicator was never wrong — it reported one-sided movement accurately every time, and it was being asked a question about exhaustion that it has no way to answer.
The second failure is acting on entry into the zone. Leaving it is the event.
A third is using the default period without deciding. 14 is a starting point.
A fourth is treating divergence as a timing tool. It can persist for a long time.
A fifth is taking a stop from the indicator. It contains no price levels.
And a sixth is reading extremes in a thin market. Very little activity produces the same number.
Related
RSI explains the calculation and why the period changes what it can see. RSI divergence is the one configuration that carries directional information. And trading range is the condition where the default reading actually applies.
The reframe that made it useful was to stop treating the levels as instructions. A reading of 78 tells me buyers have had it their own way for a while. Whether that means exhaustion or strength depends entirely on what kind of market I am in — and deciding that comes before I look at the indicator at all.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.