Relative Strength Index (RSI): how to read and use it

The Relative Strength Index (RSI) is a momentum oscillator. It condenses the recent balance of up days and down days into one number between 0 and 100. J. Welles Wilder Jr. introduced it in 1978 in New Concepts in Technical Trading Systems, the same book that gave traders the Average True Range, the Parabolic SAR and the Directional Movement System. Nearly fifty years later RSI is the most widely used oscillator on retail charting platforms. It is also one of the most misread.

This guide explains exactly what RSI calculates, how to compute it by hand, the four ways traders read it, why "overbought" is not a sell signal on its own, and how Finabase computes and scores it.

What RSI measures

Price alone tells you where a market is. Momentum tells you how forcefully it got there. RSI asks a simple question about the last n periods: of all the movement, how much of it was upward?

If every one of the last 14 closes was higher than the one before, there were no losses and RSI is 100. If every close was lower, RSI is 0. If the average gain and the average loss were the same size, RSI is exactly 50. Everything else falls in between, so RSI works on any stock, ETF, currency or coin at any price level. That is a big reason it caught on: a reading of 75 means the same thing on a $20 stock as on a $2,000 one.

RSI has nothing to do with "relative strength" in the sense of comparing one stock with another or with the S&P 500. That comparison is usually called relative strength or a price relative (for example AAPL ÷ SPY). Wilder's RSI looks at one instrument against its own recent history.

The formula, step by step

RSI is built in three steps: measure each period's change, average the gains and the losses separately, then turn the ratio of those averages into a 0–100 scale.

Two details matter more than they look.

Losses are counted as positive numbers. A close that falls $1.50 adds a loss of 1.50, not −1.50. Average gain and average loss are therefore both positive, and RS is always zero or greater.

The averages are smoothed, not simple. After the first value, Wilder does not drop the oldest period and add the newest, as a simple moving average would. He keeps 13/14ths of yesterday's average and adds 1/14th of today's value. This is an exponential moving average with a smoothing factor of 1/n (equivalent to a standard EMA of length 2n − 1). Every past price keeps a small, fading influence. That is why RSI values start out unstable: different platforms that begin the calculation on different dates show slightly different RSI numbers for the first few dozen bars, and then converge.

A worked example

Suppose that after the first 14 days a stock's average gain is $1.20 and its average loss is $0.80.

On day 15 the stock closes $2.00 higher. That is a gain of 2.00 and a loss of 0.

A strong up day moved RSI only from 60.0 to 62.9. The smoothing is deliberate. It keeps one outsized day from whipsawing the reading. It also means RSI reaching 80 or 20 takes a sustained run in one direction, not a single shock.

Reading RSI: the four main techniques

1. Overbought and oversold levels

Wilder suggested 70 and 30 as the boundaries of normal behavior. Above 70, gains have heavily outweighed losses over the look-back window. Below 30, losses have dominated. Many traders treat a move back through the line (falling below 70 or rising above 30) as more meaningful than the first touch. The crossing shows that the one-sided pressure is easing, not just that it exists.

The chart above shows why the levels are not signals by themselves. Apple's RSI hit its lowest reading of the year, 17.9, on January 20, 2026, at a close of $246.70. Over the next 20 sessions the stock rose about 7%. An "oversold" buyer who bought there would have been happy. But in May 2026 RSI went above 70 and stayed there for 15 consecutive sessions (May 8–29). Over that stretch the price climbed from $293 to $312. Anyone who sold or shorted on the first overbought reading gave up the whole move. The top came only after RSI had made its high of 79.8 on May 28 and turned down.

Adjusting the levels. Highly volatile instruments such as small caps and crypto spend more time at extremes, so some traders use 80/20 for them. Slow-moving, low-volatility instruments rarely reach 70 or 30, and 65/35 may fit them better. If you change the levels, change them for a reason you can state, and test the change rather than tuning it until a past chart looks perfect.

2. The 50 line and range shifts

The midpoint is underrated. When average gains exceed average losses, RSI is above 50. A stock in a healthy uptrend tends to hold above 40–50 on pullbacks, and its rallies push into the 70s. In a downtrend the pattern flips: rallies stall around 50–60 and declines reach into the 20s.

Technician Andrew Cardwell popularized this "range shift" view in the 1990s. In bull phases RSI tends to oscillate roughly between 40 and 80. In bear phases it tends to sit between 20 and 60. A practical reading follows from it. In an uptrend, a dip to 40–45 that holds is often a better entry than waiting for an "oversold" 30 that may never come. A stock whose RSI suddenly can't get above 60 on rallies may be shifting from a bull range to a bear range.

3. Divergence

A divergence happens when price and RSI disagree.

• Bullish divergence: price makes a lower low, but RSI makes a higher low. The second sell-off reached a lower price with less downside momentum than the first.
• Bearish divergence: price makes a higher high, but RSI makes a lower high. The advance is continuing on fading momentum.

The Apple data includes a clean real example. On March 30, 2026 Apple closed at $246.63, a hair below its January 20 close of $246.70. That was a lower closing low. But RSI on March 30 was only 35.5, far above January's 17.9. The selling that pushed price back to the same level carried much less momentum. Over the following 20 sessions the stock rose almost 10%.

Divergences have a real weakness: they can keep building. A stock can post three or four bearish divergences while it keeps rising. Treat divergence as a warning that the trend is ageing, and wait for price itself to confirm a change, for example by breaking a trendline, a recent swing low, or a moving average.

4. Failure swings

Wilder himself put more weight on failure swings than on divergences. A failure swing is defined entirely within RSI, with no reference to price:

• Bottom failure swing (bullish): RSI falls below 30, bounces, pulls back but stays above 30, then rises above its bounce high.
• Top failure swing (bearish): RSI rises above 70, drops, rallies but fails to get back above 70, then falls below its previous dip.

The logic: the first extreme shows exhaustion, the failure to reach it again shows the pressure is gone, and the break of the intermediate RSI swing is the trigger.

Choosing the period

Fourteen is a convention, not a law. Wilder used it for daily commodity charts. The period controls the trade-off between sensitivity and reliability:

The 2-period RSI is a well-known special case. Larry Connors and Cesar Alvarez published rules in 2008 that buy pullbacks when a stock above its 200-day moving average has a 2-period RSI under 10, then exit on strength. It works on a different idea from the classic 70/30 reading. It isn't looking for a reversal of the trend. It's looking for a short-term dip inside a trend you've already confirmed some other way.

A shorter period on a longer timeframe and a longer period on a shorter timeframe can describe similar spans. A 14-period RSI on a weekly chart covers about three months. When you compare readings, compare like with like.

Where RSI fails, and how to protect yourself

Strong trends. As the Apple example shows, the classic levels fail in exactly the markets where price moves most. One defense is a trend filter: only act on oversold readings when price is above a rising long-term average (such as the 200-day simple moving average), and only act on overbought readings when price is below a falling one.

Gaps and news. RSI uses closes. An overnight gap on earnings or a guidance change can throw the reading to an extreme in one bar for reasons no momentum model captures. Check the calendar before you read too much into a spike.

Thin markets. Illiquid stocks with erratic closes produce erratic RSI. The indicator assumes the close is a fair summary of the session.

Curve-fitting. It is easy to find a period and pair of levels that would have caught every turn in one stock's last year. That tells you very little about the next year. If you test RSI rules, test them across many symbols and periods, and keep the rules simple.

RSI compared with other oscillators

How Finabase calculates and scores RSI

Finabase computes RSI with the same open-source TA-Lib routines that power many professional platforms, using Wilder's smoothing on daily (or intraday) closes. The default period is 14 and you can change it on any chart.

On the symbol page's technical summary and in the Scanner, each indicator contributes a reading to a Buy / Neutral / Sell gauge. For RSI the rule is the classic one:

• RSI above 70 → flagged Overbought, counted on the bearish side.
• RSI below 30 → flagged Oversold, counted on the bullish side.
• Anything between → neutral.

Read the gauge as a summary of where each indicator stands, not as a recommendation. As this guide shows, an Overbought RSI in a powerful uptrend is often a sign of strength. Use the Scanner to find candidates, then open the chart and read the context: the trend, the 50 line, any divergence.

Frequently asked questions

Is an RSI of 70 a sell signal?

Not on its own. It means gains have outweighed losses heavily over the last 14 periods. In a trending market that can last for weeks. Many traders wait for RSI to fall back below 70, for a bearish divergence, or for price to break support before acting.

What is the best RSI setting?

There isn't one. Fourteen is the standard and a sensible starting point. Shorter periods suit short-term mean-reversion systems and longer ones suit position trading. Whatever you choose, test it across many symbols rather than tuning it to one chart.

Why does my RSI differ from another website's?

Usually because of the averaging method (Wilder smoothing versus simple averages), the amount of history used to start the calculation, or the price used (some platforms use adjusted closes that include dividends and splits). Small differences in the first bars of a chart are normal.

Does RSI work for crypto and forex?

The math works on any price series. Crypto trades around the clock and is more volatile, so readings reach extremes more often and many traders widen the levels to 80/20. Forex majors are less volatile, and 70/30 or tighter levels are common.

Summary

RSI is a compact, well-designed measure of momentum. Its strength is telling you what kind of market you are in: which side of 50 it holds, how far its swings reach, and whether momentum is confirming or contradicting price. Its weakness is the one most people learn first, the idea that 70 means sell and 30 means buy. Learn to read the context, combine RSI with an independent input, and test any rule before you trust it.