RSI Indicator Explained: Momentum, Divergence and Limits
The relative strength index (RSI) is a bounded momentum oscillator derived from average gains and losses over a defined number of price observations. A commonly used 14-period RSI ranges from 0 to 100 and is often discussed using 70 and 30 reference values. Those thresholds describe the indicator, not a guarantee that a price is about to reverse.
How RSI is derived
In Wilder’s method, measure the magnitude of positive and negative closing-price changes over the chosen period. Calculate smoothed average gains and average losses, then relative strength RS = average gain / average loss. RSI = 100−100/(1+RS). Initial smoothing and later recursive updates are important: the calculation is not simply the ratio of up days to down days. Implementations may vary slightly in how averages are initialised.
Worked example using smoothed averages
Suppose a fictional 14-period calculation produces an average gain of 0.80 price units and average loss of 0.40 units. RS = 0.80/0.40=2. RSI = 100−100/(1+2)=66.67. That number means gains have been larger on average than losses according to the specific smoothing method; it is not a 66.67% probability of an up day or a predicted return.
What overbought and oversold mean
Market commentary often labels RSI above 70 ‘overbought’ and below 30 ‘oversold’. In sustained trends, RSI can remain elevated or depressed for extended periods as prices continue moving in the same direction. A high value can signify strong recent positive momentum instead of imminent downside. The terms are historical-indicator descriptions rather than proof of fundamental overvaluation or undervaluation.
Bullish and bearish divergences
A bearish divergence may be described when price makes a higher high while RSI makes a lower high; a bullish divergence reverses that pattern near lows. These relationships can highlight slowing momentum, but identifying the relevant peaks often requires subjective judgment. If a swing high is confirmed only after subsequent bars, a historical chart can create look-ahead bias. Define peaks and confirmation delays in advance if evaluating predictive value.
Trend and range contexts
Oscillators can behave differently in a strong trend than in a sideways market. A trending instrument may show repeated high RSI readings because its gains dominate recent losses; in a range, the indicator may oscillate more often near conventional thresholds. Without clearly defining market regime and exit conditions, simple 70/30 rules can produce long runs of losses. No fixed threshold is universally optimal.
Sensitivity to period length
Shorter RSI windows typically react more quickly to recent changes but can be more volatile. Longer windows smooth the indicator and may reduce signal frequency. The same RSI period on daily and intraday charts describes different horizons. Do not compare readings across markets without checking input prices, trading sessions and smoothing conventions.
RSI and risk management
RSI says little directly about executable spread, order-book depth, fundamental catalysts or maximum possible loss. A hypothetical strategy buying solely because RSI crosses below 30 can suffer losses during a major downtrend. Any research design should specify trade timing, costs, position size and stop execution assumptions independently from the oscillator.
Evaluate RSI claims carefully
Reproduce calculations on a small dataset to confirm platform consistency. Define signals before testing, reserve out-of-sample data and compare against simple momentum benchmarks. Inspect losing regimes rather than collecting successful chart screenshots. If a divergence strategy has no measured after-cost advantage, treat the indicator as descriptive context, not evidence of a reliable edge.
For further learning on this method and its limitations, see technical indicators and RSI interpretation.
RSI: relative strength calculation
This educational comparison uses hypothetical figures and is not a prediction or investment recommendation.
| Input | Scenario A | Scenario B |
|---|---|---|
| Smoothed average gain | 1.2 | 2.0 |
| Smoothed average loss | 0.8 | 0.5 |
| Relative strength (gain ÷ loss) | 1.5 | 4.0 |
| RSI = 100 − 100 ÷ (1 + RS) | 60 | 80 |
| Meaning | Historically stronger gains than losses | Stronger gain/loss ratio, often described as overbought |
An RSI of 80 is not an 80% probability that a market will reverse, just as RSI 20 does not make a rebound statistically guaranteed. It describes recent relative gains and losses under a defined smoothing method. A trend can maintain elevated readings for many periods, and an indicator may be sensitive to the closing-price series, timeframe and warm-up. Compare the RSI with the market regime and independent evidence rather than treating a numerical threshold as a standalone trading instruction.
RSI calculation: a reproducible numerical example
RSI is based on the ratio of smoothed average gains to smoothed average losses over a selected number of periods. Under Wilder’s conventional formulation, RS = average gain ÷ average loss and RSI = 100 − 100 ÷ (1 + RS). Suppose the smoothed average gain is 0.60 price units and the smoothed average loss is 0.30 over the same defined interval. Then RS is 2 and the resulting RSI is approximately 66.67. The number means gains have been larger than losses under the specified smoothing procedure; it does not mean that a rally has a 66.67% probability of continuing.
| Input | Calculation | Illustrative output |
|---|---|---|
| Smoothed average gain | Given | 0.60 |
| Smoothed average loss | Given | 0.30 |
| RSI | 100 − 100 ÷ (1 + 0.60/0.30) | 66.67 |
Chart platforms may initialise smoothing differently or use different price series. The common 14-period setting is a convention, not an empirically proven optimum for every instrument. If average loss is zero and average gain is positive, platforms conventionally report RSI as 100; if both are zero, software handling can differ, so the exact implementation should be documented.
How to test an RSI strategy without misleading yourself
Start with an explicit research question, such as whether a specified RSI threshold adds predictive information after conditioning on an instrument’s trend, volatility and trading session. Define signal calculations using only data available at the decision time; if you calculate a daily RSI from the day’s close, a fill at the same closing price may be unattainable. Apply realistic spreads, market impact and fees, and test performance on data not used to choose the thresholds.
A signal below 30 is not inherently a buying opportunity, and an RSI above 70 is not inherently a sell signal. Persistent trends can hold the indicator in an extreme region for long periods. Compare the result with a simple price-momentum baseline and examine how conclusions change by instrument and market regime. See Trading Strategy Backtesting for look-ahead bias and out-of-sample assessment, and Position Sizing Explained for separating signals from risk limits.
Frequently asked questions
Does RSI above 70 mean sell?
No. Strong trends can keep RSI elevated for prolonged periods.
Is RSI a price forecast?
No. It summarises smoothed historical gains and losses.
Can RSI divergence fail?
Yes. Price can continue trending despite apparent momentum divergence.