An instrument in a powerful trend can hold an RSI above 70 for weeks while price keeps climbing. Traders who were taught that 70 means sell will have shorted it four times by then, each time at a better price for whoever was on the other side. The indicator was not wrong. The lesson attached to it was.
What the number is counting
J. Welles Wilder introduced the Relative Strength Index in his 1978 book on technical trading systems, and the arithmetic has not changed since. Over the lookback window, usually 14 periods, the calculation separates the bars that closed up from the bars that closed down. It averages the up moves and the down moves separately, using a smoothing method that carries previous averages forward. The ratio of those two averages is fed into a formula that compresses the result onto a 0 to 100 scale.
Read that description again and notice what is missing. There is no price level in it, no volume, no reference to any other instrument, and nothing about how far price has travelled in absolute terms. RSI knows only the balance between up closes and down closes inside its window. A reading of 82 means the recent bars have been overwhelmingly positive. That is the entire message.
Overbought means fast, not expensive
The vocabulary is the problem. "Overbought" sounds like a valuation judgement, as though the market has made an error that is about to be corrected. What the reading actually describes is speed: the move up has been one-sided and rapid relative to the last fourteen bars.
One-sided and rapid is exactly what a strong trend looks like. So the high readings cluster precisely where a trend is healthiest, and fading them means fighting the dominant flow. In a range, the same reading has more meaning, because a range is by definition a place where extremes revert. The indicator is identical in both cases; the context around it decides whether the signal is worth anything, which is why reading RSI without first classifying the market as trending or ranging produces contradictory results. Our note on trend behaviour covers the classification step that has to come first.
No oscillator reading is an instruction. RSI at 15 does not mean a bottom is in, and price can keep falling for as long as it likes. Any position taken on an indicator still needs a stop and a size the account can absorb, because leveraged trading carries a high risk of loss.
The 14 default, and what changing it does
Fourteen periods is a convention inherited from the original work, not an optimum. Shorten it to 7 or 5 and the line becomes volatile, touching the extremes several times a session; lengthen it to 21 or 28 and the extremes become rare events. Neither version is more accurate. You are choosing how often to be told something.
The trap is treating the threshold as fixed while the period moves. If you halve the lookback and keep 70 and 30, you have quietly multiplied the number of signals you will act on, most of them meaningless. Traders who adjust the period usually adjust the levels too, using 80 and 20 on fast settings to keep the signal count sane. Whichever combination is picked, it needs to be tested on the timeframe it will run on, and then left alone.
Range shifts: the useful reading in a trend
There is a more productive way to use RSI in a trending market, and it comes from watching where the oscillator stops falling rather than where it stops rising. In a sustained uptrend, pullbacks often bottom out with RSI somewhere in the 40s rather than reaching 30, and the oscillator then turns back up. In a sustained downtrend, rallies often stall in the 50s or 60s without reaching 70.
That asymmetry is informative. When an uptrend that has been holding a floor around 40 suddenly breaks down to 25, something in the character of the move has changed, and that is a genuine signal about participation rather than a threshold crossing. It is also fragile evidence on its own, so it works better alongside the price levels that mark where the trend structure would actually fail.
Divergence and its false-positive problem
Bearish divergence is a higher high in price accompanied by a lower high in RSI. The interpretation is that the second push was weaker, so the trend is tiring. It is the most popular RSI signal and the one that costs the most money.
The reason is arithmetic rather than psychology. Momentum almost always peaks before price in any extended move, because the fastest part of a trend tends to come early. So divergence appears routinely in healthy trends and keeps appearing, sometimes three or four times, while price continues. Traders who use divergence seriously treat it as a reason to tighten management rather than a reason to reverse, and they wait for price itself to break a swing level before acting. Divergence on the MACD histogram has the same statistical problem for the same reason, since both are momentum measures derived from recent price changes.
What RSI cannot see
It cannot see gaps as anything other than one large up or down close, so a weekend gap distorts the average for the full window length afterwards. It cannot see volatility: a 14-period window of tiny moves and a window of violent ones can produce the same reading, which is why a separate volatility measure belongs on the screen next to it. It cannot see the instrument's context, so an RSI of 30 on a currency pair in a central bank tightening cycle carries different weight from an RSI of 30 in a quiet August session.
None of that makes it a bad tool. It makes it a narrow one, and narrow tools are only dangerous when their users forget the boundary. RSI answers a single question well: how one-sided has recent price action been. Everything else on the chart still has to answer the rest.
"Traders lose money on RSI for one reason. They read the word overbought and hear expensive, when the number is only telling them the move has been fast."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- RSI compares average up closes with average down closes over its lookback, so it measures one-sidedness rather than value.
- Extreme readings cluster where trends are strongest, which is why fading 70 in a trend is a repeat offender for losses.
- Shortening the period without moving the thresholds silently multiplies the number of signals you will act on.
- Divergence appears routinely inside healthy trends, so it works better as a management prompt than as a reversal trigger.
Frequently Asked Questions
Does an RSI above 70 mean an asset is about to fall?
No. A reading above 70 says that recent gains have been large relative to recent losses over the lookback window. Strong trends can hold high readings for weeks while price continues in the same direction. Selling purely because the oscillator crossed a threshold is one of the most common and most expensive misreadings of the tool.
What RSI period should I use?
The default is 14 periods, from Wilder's original work. Shorter settings react faster and reach the extremes far more often, which suits short-horizon traders who expect many signals and accept many false ones. Longer settings smooth the line and produce fewer readings at the extremes. Changing the period changes how often 70 and 30 are touched, not what they mean.
Is RSI divergence a reliable signal?
Divergence is common and most instances resolve without a reversal, because momentum slowing is a normal feature of trends that continue. Traders who use it generally require confirmation from price structure, such as a broken swing level, and they size positions on the assumption that the signal will often fail. Trading on divergence alone carries a high risk of loss.