RSI Indicator Explained: Formula, Levels and Divergence

What RSI measures.

The Relative Strength Index, associated with J. Welles Wilder, summarises the balance between recent gains and losses in the same instrument. It does not compare that instrument with a market benchmark, despite the words “relative strength”. The commonly used lookback is 14 periods; on an hourly chart that means 14 hourly changes, not 14 days.

Fidelity: RSI definition and calculation describes the indicator and the conventional 30 and 70 reference levels. Those levels are descriptive thresholds. Price can continue falling with a low RSI, or continue rising with a high RSI, so “oversold” is not synonymous with undervalued.

Use completed observations when studying a close-based rule. RSI on the current candle moves as that candle's price changes. A momentary crossing that disappears before the close is different from a confirmed close-based crossing.

Calculate RSI and understand the smoothing.

For a conventional 14-period Wilder calculation, begin with 14 consecutive price changes, requiring 15 closes. Separate each change into a non-negative gain and a non-negative loss magnitude. The first averages are the arithmetic means of those 14 values.

For later observations, update each average as (previous average × 13 + new value) ÷ 14. This differs from recalculating a rolling simple average each time. It also differs from the usual EMA coefficient 2 ÷ (n + 1). Confirm the method used by your platform before comparing values.

If average loss is zero and average gain is positive, the limiting RSI value is 100; if gain is zero and loss is positive, it is 0. If both are zero, the formula is indeterminate, and software conventions can differ. Initialisation and the amount of warm-up history can also cause small discrepancies.

An original numerical example.

Suppose the current smoothed average gain is 1.5 price units and average loss is 0.5. RS is 3, making RSI 75. These are arbitrary price-change units, not a live instrument quote.

The next close falls by 2 units. The new gain is zero and the new loss is 2. Applying a 14-period Wilder update gives an average gain of (1.5 × 13) ÷ 14 = 1.392857 and an average loss of (0.5 × 13 + 2) ÷ 14 = 0.607143. RSI becomes approximately 69.64.

The indicator has crossed below 70 after one decline. That arithmetic does not tell us whether the next close will be higher or lower. It demonstrates how the current change interacts with the existing smoothed history, which is why a fixed price move does not always create the same RSI change.

Distinguish levels, momentum and divergence.

For a divergence study, define which completed swings are being compared and when they became identifiable. Connecting whichever two points look best after a reversal produces a persuasive picture but an unusable test.

A useful counterexample is a trend that makes three successive higher highs while RSI makes lower highs. A trader who shorts each divergence may take repeated losses before any reversal occurs. Keep that possibility in the explanation, rather than showing only the final successful turn.

Choose a question before changing the settings.

A shorter lookback responds to fewer observations; a longer lookback smooths over more history. Neither is inherently the best setting for forex, gold or a particular timeframe. Changing both the timeframe and lookback changes the question you are asking.

For an educational comparison, keep one price dataset fixed and examine 7-, 14- and 21-period RSI using the same smoothing and warm-up method. Record how frequently each crosses a threshold, how many crossings reverse promptly, and what the complete entry/exit rule would cost. Reserve a later, untouched period for evaluation.

Avoid treating MACD and RSI agreement as two independent votes. Both are transformations of price history. A separate observation about execution costs or an approaching scheduled economic release may answer a different question more usefully.

Work through it yourself.

Follow RSI, moving-average and MACD calculations using an editable practice price series.

Free to download without registering. Practice examples explain the method; they do not establish a profitable strategy.

Common RSI questions.