TL;DR: RSI compares the average gain on up periods with the average loss on down periods over a lookback, conventionally 14 periods, and converts the ratio into a 0 to 100 scale: RSI equals 100 minus 100 divided by (1 plus RS), where RS is average gain divided by average loss. With average positive closing price changes of $2 and average absolute negative changes of $1, each averaged across all periods in the lookback, RS is 2.0 and RSI is about 66.7, since 100 minus 100 divided by 3 equals about 66.7. The number summarizes the recent path's one sidedness. It contains no information about the company's earnings, assets, or worth, and treating it as if it does is the beginner's central error with every indicator.
Key Takeaways
- RSI is a momentum oscillator built only from past prices. Its inputs are closes; its output is a description of the recent path, nothing else.
- The standard settings are a 14 period lookback and the 70 and 30 levels as overbought and oversold markers. Both conventions descend from J. Welles Wilder's 1978 introduction and are conventions, not constants of nature.
- In strong trends RSI pins at extremes for long stretches: a stock can stay above 70 through a large advance, which is why selling strength mechanically on the level alone fails so reliably.
- Divergence, price making a new high while RSI makes a lower one, is the sophisticated use: it flags decelerating momentum inside a still rising price, a warning, not a verdict.
- RSI sits in research as a timing and context instrument beside the fundamental work, never in place of it, the same subordination as every filter in fundamental stock screener.
How is RSI actually computed?
Wilder's method: over the lookback, average the gains and the losses across all periods, where each period contributes its closing price gain or zero to one side and its loss magnitude or zero to the other, so flat and opposite periods still count; the first averages divide the 14 period totals by 14, and later values use Wilder's smoothing, which carries prior values forward rather than taking a simple mean. Divide average gain by average loss to get RS, then transform. The transformation's job is bounding: RS can run from zero to infinity, while RSI compresses it into a scale where 50 means gains and losses balanced, readings above mean gains dominated, and readings below mean losses dominated. At the extremes the compression matters: RS of 9 and RS of 19 both print as RSI in the 90s, hiding a doubling of dominance inside a few points of scale. The number is most informative in its middle ranges and as a series through time, watched for its pattern of highs and lows across a trend, which is the way its inventor intended it to be read.
Settings change the personality. Shorten the lookback to 7 and RSI becomes jumpy, whipping across levels on ordinary noise. Lengthen it and the series smooths into a slower trend gauge that stops pretending to time anything weekly. Neither is correct; they measure different horizons, and the honest practice is to choose the horizon that matches the decision the indicator is informing, then leave the setting alone, because re tuning settings until the chart agrees with your hope is data mining performed by hand.
What do overbought and oversold actually mean?
They mean the recent path was one sided, full stop. A stock above 70 has, over the lookback, gained persistently. Markets being what they are, persistent gainers frequently continue, because the forces producing the gains, earnings momentum, sector flows, a repricing thesis arriving in installments, rarely exhaust themselves on the indicator's schedule. The levels are better read as regime labels than as reversal signals: above 70 the stock is in a strong uptrend regime where dips tend to be bought, below 30 in a weak regime where rallies tend to be sold, and the middle is contested ground. Wilder himself used the levels as alerts to look closer, and built his trading rules around subsequent price confirmation, a nuance the one line summaries that circulate today mostly omit.
Used with that respect, RSI earns its place in a research workflow: flagging when a stock on your watchlist has pulled back within an intact uptrend to the zone where its momentum resets, or warning that a falling knife's RSI of 18 describes a regime of persistent selling that value arguments alone will not interrupt on your schedule. In both cases the indicator informs timing around a decision made on fundamentals. It does not make the decision, and a portfolio built by obeying it solo is a machine for buying weakness into bankruptcies and selling strength out of compounders, alternately, at scale.
How does RSI relate to the other tools on the page?
It belongs to the momentum family, beside moving averages, and the two check each other: price above a rising long average with RSI recovering through 50 describes a trend resuming; price at new highs with RSI diverging describes a trend tiring. Volume and beta add context of participation and market sensitivity. None of them, separately or together, answer the valuation question, which is why this site's research order stands: the business first, through its statements and earnings reports, the price second, and the path of the price, RSI's whole jurisdiction, third and last.
FAQ
What RSI period should I use? The conventional 14 matches the indicator's published design and the figures most sources print, which makes comparison possible. Adjust only with a reason tied to your decision horizon, and record the setting in your notes so your own history stays comparable.
Is RSI useful for long term investors? Marginally, as a patience instrument: it can keep an investor from chasing a vertical move or from panic reading a normal pullback as a collapse, by giving the path a number. It should never veto or trigger a long term position by itself.
What is RSI divergence in plain terms? Price pushes to a new high while RSI's corresponding peak is lower than its previous peak, or the mirror at lows. Recent advances are being made with less one sided force than earlier ones. It is a deceleration signal with a poor sense of timing: divergences can persist through substantial further moves.
Why does RSI differ between my broker and a website? Lookback settings, smoothing variants, data adjustments for splits and dividends, and the period basis, daily versus weekly, all change the figure. Use one source consistently for your own comparisons and treat small differences between sources as methodology, not error.
Sources
- J. Welles Wilder Jr., New Concepts in Technical Trading Systems (1978), the original presentation of the relative strength index.
- Investor.gov (U.S. Securities and Exchange Commission), investor education material on investment analysis, accessed October 2026.
- Every worked example in this post was computed directly from the stated inputs shown in the text.
Educational content, not financial advice. Any worked example uses hypothetical, illustrative figures chosen for round numbers. It is not a projection of any security's returns and not a recommendation to buy or sell any security.
