RSI: measuring exhaustion

The relative strength index might be the most-installed and least-understood tool in trading. Most users learn exactly two facts — above 70 means overbought, below 30 means oversold — and proceed to lose money fading every strong trend they meet. The indicator deserves better, and so does your account. RSI is genuinely informative once you know what it measures, which is not “too high” or “too low” but something closer to the recent balance of aggression between buyers and sellers.

It was designed in an era of daily charts and 14-day settings, and that default persists everywhere, including the indicator panels on modern terminals. The defaults are fine. The interpretation is where the casualties happen.

What the number actually is

Take the last 14 bars. Average the gains on the up bars; average the losses on the down bars (smoothed, in the standard formulation). RSI converts the ratio of those two averages onto a 0–100 scale. If average gains and losses are equal, RSI sits at 50. If the last 14 bars were all up, it pins near 100. A reading of 70 means recent up-moves have been a little over twice the size of recent down-moves on average. Notice what is absent from that sentence: nothing about valuation, nothing about “due for a pullback.” RSI measures one-sidedness of recent movement. That is the whole machine.

The overbought misconception

Here is the uncomfortable fact: an RSI above 70 means buyers have been dominating, and buyer domination is what strong uptrends are made of. In a powerful trend, RSI can reach 80, stay above 70 for weeks, and the “overbought” market can double afterward. Fading high RSI in a trending market is shorting strength precisely because it is strong — a strategy with an impressive record of converting traders into former traders. The mean-reversion read works best where mean reversion lives: in ranges, where an RSI extreme at the range boundary genuinely does mark the zone where the move tends to tire.

Picture the failure in slow motion. An asset breaks out of a base and starts trending; by the third week RSI touches 74 and the fade trader shorts, because the textbook said overbought. Price grinds higher; RSI holds 70-plus for another month while the position bleeds. The same logic in mirror image is just as expensive on the long side: buying every RSI 28 print in an established downtrend means catching a falling market repeatedly on the theory that it has fallen a lot — which is true, and is also exactly what downtrends do. In both cases the trader treated a description of strength or weakness as a prediction of its end. RSI never made that promise.

Divergence: the second look

Divergence is RSI’s most interesting trick. Price makes a higher high; RSI makes a lower high. The translation: the second push reached a new price, but the average size of up-moves behind it shrank — the advance is being carried by less force. Bearish divergence at highs and bullish divergence at lows both flag the same thing: momentum fading while price keeps stretching. Two honesty clauses apply. First, divergence is a warning, not a signal — strong trends routinely print two or three divergences before actually turning, and traders who short the first one fund the move’s final leg. Second, divergence tells you nothing about timing or magnitude. It earns a place as a reason to tighten stops, take partial profit, or demand more confirmation — not as a reason to step in front of a moving market.

Range shifts: RSI changes its home

The subtlest and most practical RSI behavior: the indicator occupies different territory in different regimes. In healthy uptrends, RSI tends to oscillate roughly between 40 and 80 — pullbacks bottom out near 40–50 without ever reaching “oversold.” In downtrends, the band slides to roughly 20–60, and rallies exhaust near 50–60 without reaching “overbought.” Waiting for RSI 30 to buy a dip in an uptrend means waiting for a dip deep enough that the uptrend may no longer exist. The range shift also gives you a regime clue in reverse: when an uptrend’s pullback drives RSI to 25 for the first time in months, the character of the market has changed, and the old playbook deserves an audit.

• Uptrend habitat: roughly 40–80, with pullback lows near 40–50.
• Downtrend habitat: roughly 20–60, with rally highs near 50–60.
• A decisive break out of the habitual band is regime information, not noise.

Settings and timeframes

Shorter lookbacks (7, 9) make RSI twitchier — more signals, more junk. Longer ones (21, 25) smooth it into a slow trend gauge. The standard 14 is a reasonable middle and, importantly, it is what most other participants are watching, which matters for the same coordination reasons round numbers matter. Whatever you choose, remember that RSI on a 5-minute chart and RSI on a daily chart are answering the same question about completely different crowds. An oversold scalper’s chart inside an overbought daily trend is not a contradiction; it is two timeframes doing their jobs.

Key takeaways

• RSI measures the recent ratio of average gains to average losses — one-sidedness of movement, not value.
• Overbought is not a sell signal; strong trends live above 70, and fading them is how the lesson gets expensive.
• RSI extremes are fade-worthy mainly in ranges; classify the market before applying the classic thresholds.
• Divergence flags fading momentum but routinely fires early — treat it as a reason for caution, not an entry.
• RSI shifts its operating range with the regime: roughly 40–80 in uptrends, 20–60 in downtrends, and a break of that habitat is news.