RSI indicator: settings and signals

RSI measures the ratio between the strength of gains and the strength of losses over the last N bars and normalizes it onto a 0–100 scale.

What the indicator looks like on a real chart

Try it yourself: how the period changes the signals

What the indicator actually measures

Wilder built RSI as an answer to a problem with plain momentum: the price difference over N bars scales with the price level itself and makes it impossible to compare different instruments. RSI solves this by dividing average gains by average losses, so the resulting value is dimensionless and reads the same on any asset and any timeframe.

A detail that often gets missed: the calculation uses not a simple average but Wilder smoothing — each new value shifts the accumulated average by only 1/N. Because of this RSI has a long memory: a single anomalous bar keeps influencing the indicator dozens of bars later, and after a strong impulse the value returns to 50 slowly and reluctantly.

Hence the key property: in a sustained trend RSI does not oscillate around the midpoint but pins itself to the top or bottom of the scale and stays there for weeks. This is not a malfunction but a direct consequence of the formula — if there are almost no losses, the denominator is small and the value has to be high.

Formula

RS is the ratio of average gain to average loss. When the two are equal, RS equals one and RSI returns exactly 50. When there are no losses at all, the denominator goes to zero and the indicator is 100 by definition — which is exactly what happens on pullback-free impulses.

Standard settings

Period 14 is Wilder's original value, chosen for daily bars as half of a monthly cycle. Shortening it to 7 doubles the sensitivity and the number of signals; lengthening to 21–25 gives rarer but steadier entries. The 70/30 levels are not a law but a convention: in a strong trend they are shifted to 80/40 or 60/20 so the indicator does not signal against the direction of the move.

Implementation code (Python)

def compute_rsi(close, period=14):
    delta = close.diff()
    gain = delta.clip(lower=0).rolling(period).mean()
    loss = (-delta.clip(upper=0)).rolling(period).mean()
    rs = gain / loss
    return 100 - (100 / (1 + rs))

When the indicator stops working

Common mistakes in use

• Selling only because RSI is above 70. In a trend this is the most expensive mistake: being overbought in a rising market is the norm, not an anomaly.
• Treating divergence as a standalone signal. A discrepancy between price and RSI indicates only a weakening of momentum, but momentum can weaken indefinitely without a reversal.
• Tuning the period against history until the signals look pretty. That is curve fitting: an indicator with period 14 and one with period 11 differ in noise, not in quality.
• Comparing absolute RSI values across timeframes, forgetting that daily RSI(14) and hourly RSI(14) describe completely different horizons.

Who needs this and why

RSI is useful for those who have already established the market direction by another method and are looking for a moment to enter within that direction. As a standalone signal generator it is weak, but as a "do not buy at the top of an impulse" filter it works well.

Frequently asked questions

Other «Momentum» indicators