Where RSI came from
J. Welles Wilder published the Relative Strength Index in his 1978 book New Concepts in Technical Trading Systems, the same volume that introduced ATR, ADX and the Parabolic SAR. He built RSI to solve a specific problem: raw price changes carry no scale, so momentum in one stock could not be compared with momentum in another. His oscillator compresses momentum into a 0 to 100 line, which is why it still fills chart sidebars almost fifty years later.
What the formula measures
RSI compares average gains against average losses over a lookback, classically 14 periods, and scales the ratio so that equal up and down movement reads 50. Wilder smoothed the averages his own particular way, which means old bars never fully drop out of the calculation, and the RSI value on a chart can differ slightly between platforms that restart the series at different dates. It is a momentum thermometer, not a buy or sell command. The bounded scale was the design win: RSI computed on a low priced small cap and on a major index can be compared directly, which was Wilder's point from the start.
Using 70 and 30 without misusing them
Wilder suggested 70 as overbought and 30 as oversold, and half the market still treats a crossing as an instruction to fade the move. The levels were flags, not orders. In a wide swinging stock, RSI tags both extremes weekly and fading each tag is a coin flip. In context the levels work better as inputs: an uptrend pullback that holds above 40 and then turns up from 45 is behaving very differently from a rally that dies at 55 inside a downtrend. Pulling up five years of daily bars on LannaPro makes the effective ranges of a specific instrument visible in seconds.
Why RSI stays overbought for weeks in strong trends
The most expensive lesson in RSI is shorting strength. In a genuine trend, gains pile on gains, the average gain keeps climbing, and RSI parks above 70 for weeks at a stretch. Momentum leaders do this routinely, and shorting them because the oscillator looks stretched is how accounts get run over. Constance Brown made the definitive point in her 1999 book Technical Analysis for the Trading Professional: in bull regimes RSI tends to oscillate roughly between 40 and 90, and in bear regimes between 10 and 60. Read the range the indicator is actually using, not the raw level printed today.
Divergences and failure swings
A bearish divergence appears when price prints a higher high while RSI prints a lower high, telling you the latest push bought less momentum per unit of price. Bullish divergence is the mirror image at lows. Divergences are conditions rather than clocks: they can persist across three or four successive highs before anything resolves.
The failure swing is Wilder's own, less crowded variant. RSI tops above 70, pulls back, attempts a second peak, fails to exceed the first, then breaks the low of the pullback. Say the sequence runs 76, dip to 63, second try at 71, and a slice through 63: that break is the sell signal, generated by the oscillator alone, and Wilder considered it stronger than the price-confirmed kind.
Settings and practical use
The 14-period default suits daily charts. Larry Connors popularized the 2-period RSI for a different job: catching pullbacks, with readings under 10 in stocks trading above their 200-day average marking stretches that historically snap back. LannaPro lets you pin an RSI panel under price with any period, and the productive habit is checking the oscillator's recent range before trusting any single print. Shorter lookbacks such as 5 or 7 make the line jumpier and suit short term mean reversion work, while longer ones smooth out into a regime filter. Strength is a fact, not a sell signal; sell it only with a level and a plan that survive being early.