Guides
RSI explained: overbought, oversold and common traps
Updated on · Trade AI team
Understand the RSI (Relative Strength Index): calculation, the 70/30 thresholds, divergences, and why an overbought RSI is not a sell signal on its own.
Definition
The RSI, also created by J. Welles Wilder (1978), compares the strength of recent gains with recent losses. It ranges from 0 to 100. Its classic period is 14 candles.
How it is calculated
Take the smoothed average of gains and the smoothed average of losses over the period. Their ratio gives RS; the RSI equals 100 − 100 ÷ (1 + RS). If gains dominate, the RSI rises toward 100; if losses dominate, it falls toward 0.
The 70 and 30 thresholds
Above 70 is called overbought; below 30, oversold. These words often mislead: in a strong trend, the RSI can stay above 70 for weeks while price keeps rising. Selling every time it crosses 70 means fighting the trend.
That is why many methods only use the RSI in the direction of the main trend: looking for an oversold dip to buy while the market is above its long average, for example.
What our tests show
We tested Larry Connors' “RSI 2” method (buy when RSI 2 drops below 10 above the 200 average) on BTC, ETH and SOL on 4h, with real costs, over several years. It did not pass our criteria: its profit factor stayed below 1 on all three assets. It therefore gives no signal in Trade AI.
A well-known method is not necessarily effective on a given market, with real costs. Hence the value of testing before using.
Divergences
A divergence appears when price makes a new high but the RSI does not (or the reverse). It signals possible exhaustion, not a certain reversal. Used alone, it produces many false signals.