RSI is a bounded momentum oscillator: it compares recent average gains with recent average losses and expresses the result on a scale from 0 to 100. That makes it a compact description of recent price behaviour, not a forecast and not a trade by itself (CME Group's oscillator lesson, Fidelity's RSI guide).
This is deliberately a different question from our RSI divergence explainer. Divergence compares RSI pivots with price pivots; this article is about the oscillator itself, its common readings, and the rules a test would need.
What the Relative Strength Index measures
The Relative Strength Index, or RSI, turns the relative size of recent upward and downward closes into a number between 0 and 100. Its familiar formula uses relative strength, RS, before converting it to the RSI scale (Fidelity's RSI guide, Wilder's original book).
RS = average gain / average loss
RSI = 100 - 100 / (1 + RS)The commonly displayed version uses 14 periods, a convention associated with J. Welles Wilder's original presentation. It is a default, not a law of the market (Fidelity's RSI guide, Wilder's original book).
Why 70 and 30 are conventions, not commands
Readings above 70 and below 30 are commonly described as overbought and oversold. CME Group and Fidelity both present those values as widely used reference zones, while also warning that a strong trend can leave RSI elevated or depressed for an extended period (CME Group's oscillator lesson, Fidelity's RSI guide).
That warning matters. A reading above 70 says recent gains have dominated the chosen lookback. It does not establish that a reversal is imminent. The same symmetry applies below 30: recent losses have dominated, but the price path can still continue.
| Reading rule | What it describes | What it does not settle |
|---|---|---|
| RSI above 70 | Recent gains outweigh recent losses on the selected lookback | Entry timing or a reversal |
| RSI below 30 | Recent losses outweigh recent gains on the selected lookback | A market bottom |
| RSI crossing 50 | The balance of the selected averages has changed | Whether a trend will persist |
| RSI turning | The oscillator changed direction | A complete entry and exit rule |
The labels are useful shorthand. Treating them as automatic instructions is where shorthand becomes a story.
The choices hidden inside one RSI chart
An RSI line looks objective because the platform calculates it. The trading method around it may not be.
A shorter lookback reacts more quickly to recent price changes; a longer lookback smooths those changes. That follows directly from the calculation, but it does not prove either setting is better. The market, timeframe, cost model, entry trigger and exit rule all change the result.
The same applies to thresholds. A rule based on 70 and 30 is different from a rule based on 80 and 20, a 50-line cross, or no fixed threshold at all. Once a trader tries several versions and retains only the most attractive chart, the result has a selection problem rather than a validated rule. The broader evidence on technical analysis is mixed, and reviews flag data snooping, ex-post rule selection and transaction costs as recurring research problems (Park and Irwin's review, University of Illinois publication record).
How an RSI rule becomes testable
An RSI strategy is testable only when the chart annotation becomes an instruction set that a person or program could apply before the outcome is known.
| Part of the rule | Question to lock before testing |
|---|---|
| Instrument and timeframe | Which market and bar interval generate the RSI? |
| Calculation | Which lookback and smoothing method are used? |
| Signal | Is it a threshold touch, a cross, a turn, or a combination? |
| Entry | Which price and which bar permit an entry? |
| Invalidation | What ends the idea when it is wrong? |
| Exit | Is the exit time-based, price-based, or the opposite signal? |
| Costs | Which spread, commission, slippage and swap assumptions apply? |
Test a fixed RSI rule against a simple baseline rather than congratulating the indicator for a chart selected after the fact. Split the research into in-sample design and genuinely unseen data, then change one setting at a time around the chosen parameters. Out-of-sample testing and parameter sensitivity are useful checks because they ask whether the rule survives data and settings it did not get to choose.
realbacktesting is a trading-software studio for cTrader that treats testability as a design requirement. The relevant standard here is simple: define the RSI rule, include plausible execution costs, and inspect the result beyond the neatest equity curve. How to verify a cTrader backtest explains that discipline in more detail.
Frequently asked
Is RSI a buy or sell signal?
No. RSI is an oscillator derived from recent gains and losses. A tradeable method still needs a pre-defined entry, invalidation, exit and cost model.
Does RSI above 70 mean price will fall?
No. Above 70 is a conventional overbought reference, not proof of an imminent decline. Strong trends can keep RSI in an elevated zone for extended periods (CME Group's oscillator lesson, Fidelity's RSI guide).
Is a 14-period RSI always best?
No. Fourteen periods is the familiar Wilder convention. Whether another setting improves a complete rule is a question for a pre-defined test, not a visual preference.
Takeaway
RSI can describe momentum cleanly; only a rule that survives unseen data and realistic costs can earn the word edge.