The relative strength index
What it calculates
The relative strength index compares the size of recent gains with the size of recent losses and expresses the result on a scale from 0 to 100. The standard setting, and the one used on the charts here, looks back over fourteen periods.
Roughly: if every one of the last fourteen sessions was up, RSI is near 100; if every one was down, near 0; if gains and losses have been about equal in size, near 50. It is a measure of the recent balance of momentum and nothing more.
Overbought does not mean expensive
Readings above 70 are conventionally called overbought and below 30 oversold. Both terms are misleading, because neither says anything about value. An RSI of 80 means the price has risen persistently and with conviction. That is a description of what happened, not a claim that it has gone too far.
The distinction matters because the two readings are frequently the opposite of a warning. Strong uptrends routinely hold RSI above 70 for weeks, and treating each such reading as a signal to act means fighting a trend at every step of it.
It fails where it is most often used
RSI is built for range-bound conditions, where a price oscillates around a level and extremes tend to revert. In a sustained trend, which is exactly when people go looking for a reason to act, it saturates and stops discriminating: an indicator pinned near 80 for a month is telling you the same thing every day.
It is also derived entirely from price. It knows nothing about earnings, cash flow, an acquisition or a product failure. A company can be at RSI 25 because it is genuinely deteriorating, and the indicator has no way to tell you which kind of decline you are looking at.
How practitioners actually read it
The more informative use is divergence rather than level: price making a new high while RSI does not is a sign the move is being driven by fewer and smaller advances than the one before it. That is a genuine observation about market internals, though far from a reliable one.
The fourteen-period default is a convention, not a law. Shorter settings react faster and produce far more signals, most of them noise; longer ones lag. Changing the period changes what counts as a signal, which is a good reason to be sceptical of any rule stated in terms of a specific threshold.
