RSI explained: the formula, 70/30 and false signals

Relative Strength Index (RSI) turns recent upward and downward closing-price changes into a number between 0 and 100. It measures the balance of recent moves, not the probability of a reversal. Reading a high number as a compulsory sell order is the common mistake.

What goes into the calculation?

A conventional RSI uses 14 completed periods, although the interval and smoothing method can vary. Compute the average upward change and the average absolute downward change over the lookback, then form RS = average gain ÷ average loss; RSI = 100 − 100 ÷ (1 + RS). If those averages were 1.5 and 0.5 units in a simplified illustration, RS would be 3 and RSI would be 75. That is arithmetic on past closes, not evidence the next close must fall. If there are no losses in the lookback, implementations handle the division-by-zero case as a very high/100 reading; data vendors can differ at the start of a series.

The Fidelity RSI reference describes the conventional 70/30 thresholds and notes that strong trends can keep RSI elevated or depressed for extended periods. Thresholds are conventions, not physical boundaries or universal signals.

What 70, 30 and divergence actually describe

• Above 70: recent upward closes have dominated by this calculation. 'Overbought' does not tell you when a price will stop rising.
• Below 30: recent downward closes dominate. 'Oversold' is not evidence that a bounce must start now.
• Divergence: price makes a new extreme while RSI does not. This describes weaker relative momentum across the chosen swing points. It can persist or be drawn differently after the fact.
• 50: average gains and losses are approximately balanced for the selected smoothing and period. It is not a break-even return line.

An observation, not a trading signal

Suppose a share closes higher for many sessions and its 14-period RSI reaches 76. The statement you can support is 'gains dominated recent closes under this calculation'. The statements 'it has to fall now', 'a short is safe' or 'the target is X' do not follow. A price trend can continue while RSI remains high. Compare the environment, order execution and fees before evaluating any written rule; see the regime guide and the cost guide.

How to test an RSI rule without hindsight

1. Define instrument, session, bar interval, lookback, price source and the RSI implementation before inspecting results.
2. Use the RSI from a completed bar if the action would happen on the next bar. Avoid using its final close to claim an earlier fill.
3. State an entry, invalidation, position-sizing assumption and exit. A threshold by itself is not a full test.
4. Include trading costs, missed fills and different market conditions. Compare to a simple baseline and leave an untouched period for evaluation.
5. Report failed variants as well as attractive ones. Our backtesting primer and open research notes show why attractive in-sample results can disappear.

RSI and Stochastic are both momentum-family measures and can echo the same information. The indicator family guide helps decide which question an indicator answers before a chart fills up with lines.

Sources and further reading

Links were reviewed 2026-09-24. Regulatory permissions, firm status and product terms can change; use the current official register before acting.

1. Fidelity — Relative Strength Index reference

Keep learning

• Indicator families
• SMA vs EMA: what moving averages actually measure
• ATR explained: measuring movement without predicting direction