Relative Strength Index (RSI)

Reviewed 11 August 2026 · Sourced from Wilder’s 1978 book, his own 1986 article, and current platform documentation

What RSI is actually measuring

You pull up a chart. The stock closed higher on nine of the last fourteen days, and the up days look bigger than the down days. Is that a lot? Compared with what? Your eye is a bad judge of this. RSI turns the impression into a number.

It takes a fixed window of recent bars, adds up what price gained on the up bars and what it lost on the down bars, and compares the two. The result is plotted as one line beneath the price chart, on a scale that cannot go below 0 or above 100.

The name is the biggest source of confusion here. “Relative strength” does not mean relative to anything else. Not the S&P 500, not the sector, not a competitor. The only inputs are that one security’s own up closes and down closes. A stock can be down 30% for the year, losing badly to every peer it has, and still print a high RSI, because the last fourteen bars happened to be mostly green.

RSI comes from J. Welles Wilder Jr., who introduced it in his 1978 book New Concepts in Technical Trading Systems (Trend Research, Greensboro NC) as Section VI. He wrote about it again under his own byline in Technical Analysis of STOCKS & COMMODITIES in December 1986, and that article anchors several claims below.

The formula, and why the scale stops at 0 and 100

You will never type this in by hand, but its shape explains the two things people find strangest about RSI.

First: the bounds are arithmetic, not a design decision. An average gain and an average loss are both sizes, so neither can be negative, so RS can never be negative. Feed in RS = 0, a window with no up closes at all, and you get 100 − 100, which is 0. Push RS toward infinity, no down closes at all, and the subtracted term shrinks toward nothing, so the line approaches 100 without arriving.

Second: the lookback is an average, and averages are slow. With n set to 14, each new bar is one of fourteen inputs. Shorten the window and each bar carries more weight, so the line jumps around more. Lengthen it and the line smooths out.

On the number 14, be careful what you claim. 14 is documented as the default — TradingView’s official documentation states that 14 bars is the default, and StockCharts’ ChartSchool gives 14 periods. That is verifiable today. The further claim that Wilder personally chose 14 days is in almost every article written about RSI and confirmed by no primary source reachable for this page; his Section VI text sits behind a lending restriction. So: 14 has been the standard since Wilder’s work and it is what your platform loads. Who picked it, this page cannot tell you.

70 and 30 — the one pair with a source behind it

Almost everyone meets RSI through two numbers: above 70 is “overbought,” below 30 is “oversold.” Of every threshold pair attached to this indicator, that is the one you can trace.

It comes from Wilder. The published summary of his own article on the Relative Strength Index states that tops and bottoms are indicated when RSI exceeds 70 or drops below 30. That is not folklore grown up around the tool afterwards; it is the creator’s own published treatment. Both platforms checked for this page ship it as the default: Schwab’s thinkorswim study loads 30 oversold and 70 overbought, and TradingView attributes the same pair to the 1978 book.

The words mean less than they sound like they mean. “Overbought” describes the last fourteen bars. It does not forecast the next one. An RSI of 74 says gains outweighed losses over the window by enough to push the ratio past that mark. It measures something that already happened.

The threshold numbers nobody publishes

Past 70/30, RSI collects level sets the way old code collects comments. You will see them in tables and in courses, this site’s own included: 75 and 25, bands that shift with the trend, the 50 line as a trend divider. Here is what those tables leave out. No exchange, no regulator and no standard-setting body publishes any RSI threshold at all. Each one below is a rough rule of thumb many traders use, and that is the whole of its standing.

The Brown row is the useful one. Two respected sources — a charting platform’s reference school and the professional body for technical analysts — give the same author’s work different figures. Somebody has it wrong, and without the book neither this page nor the article you read last week can say who. The qualitative idea survives the disagreement: in a sustained uptrend RSI readings tend to sit higher in the range, in a sustained downtrend lower, so a fixed 70/30 pair behaves differently depending on which you are in. The exact band edges are the part to hold loosely.

Wilder’s own list of uses — including the one the course skips

Overbought and oversold is the use everyone knows. It was one of five. The published summary of Wilder’s article lists the set:

• Tops and bottoms, indicated above 70 and below 30.
• Chart formations on the RSI line itself — shapes that appear in the indicator without appearing on the bar chart.
• Failure swings.
• Support and resistance that shows up more clearly on RSI than on price.
• Divergence between RSI and price action, signalling potential turning points.

Divergence is the one most traders settle on. Price grinds out a higher high, but RSI’s peak this time is lower than its last one. Price went further; the momentum behind it did not. This is genuinely Wilder’s, listed under his own name.

Now the limit, because it matters more than the pattern does. The attribution confirms the definition and the origin. It confirms nothing about accuracy. No primary source reachable for this page puts a hit rate on RSI divergence — not how often it precedes a reversal, not how often it fails, not over what horizon. Anyone quoting you a percentage got it somewhere other than Wilder.

The failure swing is one of the five, and the course module on RSI never names it. It is named here because his published list names it. The working definition lives inside Section VI, which was not reachable, so it stays named rather than defined — a gap this page would rather show than paper over. For levels drawn on the indicator line, see the live technical page.

What a reading does not tell you

Three things get loaded onto RSI that it cannot carry.

It knows nothing about the business. The only inputs are up closes and down closes. Earnings, debt, margins, management, the lawsuit filed this morning — none of it is in there. A company can be deteriorating and print a rising RSI, because the line summarizes trading, not the company. That work happens on the fundamentals side.

It does not tell you which direction to take the signal. This is not a hedge, it is a live disagreement. Wilder’s framework reads an extreme as exhaustion. StockCharts credits Andrew Cardwell with the opposite reading — positive and negative reversals, and a reinterpretation of divergence as behaving differently in bull markets than in bear markets, inverting Wilder’s framework. Two credible readings of the same line pointing opposite ways, and no published hit rate for either — reason enough to distrust anyone who tells you what a reading “means” without saying whose framework they are using.

It does not size your risk. RSI has no opinion about how much money should be behind the trade. That question belongs to the behavior side of the market.

Where RSI sits among everything else on the chart

The instinct after learning one oscillator is to add three more. Resist it, for a structural reason: tools built on similar logic tend to say similar things, so stacking them repeats one opinion in two typefaces while feeling like confirmation. One tool per category is the stronger arrangement — something reading trend, something reading momentum, something reading volume.

RSI’s slot is the momentum one. The bounded scale makes readings comparable across securities, and the 14-period default means your chart is the chart most other people are looking at. Convention has real effects when enough participants share it, even when nothing official stands behind the numbers.

Indicators in full are Stage 4 of the Technical Analysis course. Live readings across the index sit on the technical markets page, no account required.

What trips people up

• Reading the name literally. RSI compares a security only to its own recent past, not to other securities.
• Treating 70 as a sell and 30 as a buy. Those are the levels Wilder published for indicating tops and bottoms. Nothing in his published work turns a reading into an instruction.
• Quoting the other numbers as rules. 75/25, the shifting bull and bear bands, the 50 line — conventions, all of them, and no standard-setting body publishes one.
• Assuming an extreme reading has to correct soon. The formula puts no time limit on how long the line can stay above 70 or below 30.
• Believing a divergence hit rate. Divergence is Wilder’s and it is real as a defined pattern. Any percentage attached to how often it works came from somewhere other than a primary source.
• Changing the lookback until the signal looks good. A shorter window makes the line more reactive and produces more crossings of any level. That is arithmetic, not insight.
• Using it on the money that pays rent. No indicator changes what a loss does to a household with no buffer underneath it.

The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.