What the RSI is and how to read it in crypto

The RSI is the most widely used momentum oscillator in technical analysis. Read well, it measures the exhaustion of a move; read poorly, it is the single most common cause of premature counter-trend entries.

Definition and formula

The Relative Strength Index, created by J. Welles Wilder in 1978, is an oscillator bounded between 0 and 100 that compares the magnitude of recent gains with that of recent losses. Its standard period is 14 candles.

It is worth pausing on what it does not measure. The RSI compares the asset with no other asset, despite what its name suggests: the "relative strength" is the asset against itself, between its own recent gains and losses. It also measures no volume, no trend and no volatility. It is a single dimension of the market, and that limitation explains nearly every mistake made with it.

Wilder smoothing

Wilder used neither a simple nor a conventional exponential average, but his own smoothing in which each new value weights the accumulated history. The practical effect is that the RSI remembers more than an oscillator built on fast averages: it reacts with some lag and does not lurch on a single unusual candle.

That technical detail has a very concrete consequence: two platforms implementing the smoothing differently will show different values for the same candle. Before comparing one tool's reading with another's, or with someone else's analysis, check that both use Wilder smoothing on closing prices.

The classic levels: 70 and 30

By convention, an RSI above 70 indicates statistical overbought and an RSI below 30, oversold. The 45–55 zone is considered neutral or trending.

Why overbought does not mean sell

Here is the nuance that separates the analyst from the beginner. "Overbought" describes a statistical state —price has risen a lot and fast relative to its own recent history— not an instruction. In a strong uptrend, common enough in crypto, the RSI can hold above 70 for weeks while price keeps rising. Selling on that basis is the equivalent of getting off the train because it is going fast.

The logic is clearer in reverse: a high RSI is precisely what you would expect from a trend that is working. If price rises steadily, average gains exceed average losses and the oscillator stays high by construction. The RSI only warns of something when that relationship breaks.

The ranges that reveal the trend

In trending markets it is far more useful to read the range the RSI respects than its individual readings. In an uptrend, pullbacks tend to stall in the 40–50 zone rather than reach 30, and the oscillator moves roughly between 40 and 90. In a downtrend the mirror holds: bounces die at 50–60 without reaching 70, and the range shifts to the 10–60 band.

That shift of range is itself a signal. When a market that had spent months respecting a floor at 40 starts piercing it and capping at 60, the regime has changed, even if price has not yet made that obvious.

The RSI changes with the timeframe

An RSI of 75 on 1-hour candles is everyday noise; an RSI of 75 on weekly candles describes a state of the whole cycle. The same figure means radically different things depending on the candle it is computed on, and that is the most common source of confusion for beginners.

That is why serious analysis contrasts several timeframes at once and applies a strict hierarchy: the higher timeframe rules the lower one. An oversold reading on 1 hour inside a weekly downtrend is an invitation to lose money. On Bitcoin, a weekly RSI above 85 has historically coincided with late-cycle euphoria phases, while monthly readings below 40 have only appeared at the bottoms of deep bear markets.

Divergences: the highest-quality signal

A divergence appears when price and indicator contradict each other, and it is one of the most reliable exhaustion signals in technical analysis:

Bearish divergence: price makes a higher high, but the RSI makes a lower high. The rally is losing internal strength and the probability of a top rises. Bullish divergence: price makes a lower low, but the RSI draws a higher low: selling pressure is exhausting.

How to filter out false divergences

The practical challenge is telling real divergences from noise, because on any chart you can draw dozens of them if you pick your points generously enough. Three filters cut the problem down:

• Valid pivots. The highs or lows being compared must stand out over several candles on each side, not be any minor swing.
• Reasonable distance. Two pivots right next to each other do not form a meaningful divergence; nor do two separated by half a year of a different market.
• Consistency with capital flow. A bullish divergence with no sign of accumulation in the volume indicators is usually a trap.

And a warning that saves money: a divergence is not an entry signal, it is a warning of exhaustion. Price may take weeks to respond, or never respond at all. It is for reducing exposure and tightening risk, not for opening a position against the trend.

Hidden divergences

There is a lesser-known variant that works the other way round: it confirms continuation instead of anticipating a reversal. In an uptrend, a hidden divergence appears when price makes a higher low while the RSI makes a lower one. It means the pullback has discharged momentum without breaking price structure: the market took a breather, it did not turn.

They are harder to spot and only make sense read in favour of the dominant trend. Using them against it multiplies false signals.

The RSI applied to each asset

The same indicator performs differently depending on where it is applied, and that depends above all on liquidity:

On Bitcoin, being the most liquid asset with the longest history in the market, readings are clean and ranges are respected with notable consistency. It is where the RSI works best and where it makes sense to calibrate your intuition.

On Ethereum it is worth applying twice: on the dollar chart and on the ETH/BTC ratio. An oversold RSI on the ratio does not say ETH is cheap, it says its punishment relative to Bitcoin has run a long way. Two different questions deserve two readings.

On XRP caution is required. Its pattern of long ranges interrupted by vertical moves means the oscillator saturates for days during impulses and stops informing. Trading against the move on an RSI reading is, on that asset, particularly expensive.

StochRSI: the hypersensitive cousin

The StochRSI applies the stochastic formula to the RSI itself (typical parameters 3,3,14). That is, it measures the position of the RSI within its own recent range, not that of price. The result is an ultra-fast oscillator that flags microscopic turns.

Its usefulness lies in refining entry timing inside a trend already confirmed by other means. As a standalone signal it is far too noisy: it spends most of its time pinned to one of its two extremes and produces constant crossovers that describe nothing. It should never replace the RSI, only complement it once the bigger framework is settled.

How to combine it with other indicators

The RSI measures momentum and nothing else. The read gains robustness when paired with indicators covering the dimensions it lacks: a trend one, such as Supertrend, to know which side of the market to be on, and a volatility one, such as Bollinger Bands, to know whether the move is wide or compressed.

With the MACD the relationship is direct complement: the RSI bounds the state on a fixed 0-100 scale, while the MACD describes acceleration without limits. When both show divergence at once, the signal has historically been of higher quality than either alone. What adds nothing is stacking three momentum oscillators: that is not three confirmations, it is the same data point repeated three times.

Common mistakes

• Selling just because the RSI touches 70 in a full uptrend, ignoring that this is exactly the expected behaviour of a healthy trend.
• Trading divergences on minute timeframes, where noise dominates and almost any drawing looks like a divergence.
• Ignoring the timeframe hierarchy. Daily and weekly RSI can say opposite things, and the higher one always rules.
• Comparing values across platforms without checking they use the same smoothing.
• Treating a divergence as an entry. It is a warning of exhaustion; price may take weeks to respond.

Frequently asked questions