What Is RSI (Relative Strength Index), and How Do Traders Use It?

Relative Strength Index (RSI) is a single number, bounded between 0 and 100, that measures how much of a stock's recent movement has been up versus down over a lookback window — a high reading means gains have dominated losses lately, a low reading means the opposite, and the standard read treats the far ends of that range as "overbought" and "oversold." It says nothing about direction on its own; it says how lopsided the recent push has been.

That's the whole answer. The rest of this page is how the number is actually built, what "overbought" really means versus what people assume it means, and the specific way traders misuse it that costs the most.

How is RSI actually calculated?

RSI starts from something simple: over a lookback window (14 sessions is the standard default, same as ATR's default), separate every day's price change into either a gain or a loss. Add up the average gain and the average loss over that window, then compare them — the more the average gain dominates the average loss, the higher the reading; the more losses dominate, the lower it goes. The math folds that ratio into a 0–100 scale so every stock, at every price, reads on the same bounded axis.

Two things follow directly from that construction, and both matter more than the exact formula:

1. RSI only cares about the size and frequency of up moves versus down moves, not the price level itself. A low-priced stock and a much higher-priced one can carry the identical RSI reading if their recent up/down mix has been the same shape.
2. It's bounded. Unlike ATR, which can climb without limit as a stock gets wilder, RSI cannot exceed 100 or fall below 0 — that ceiling and floor are what make "overbought" and "oversold" a meaningful vocabulary in the first place.

What do "overbought" and "oversold" actually mean?

This is the most misread part of the indicator, so it's worth being precise. A high RSI reading does not mean a stock is expensive, and a low reading does not mean it's cheap. RSI carries no information about valuation at all — it means the recent run of gains has been large and persistent relative to the losses in the same window. That's a statement about the shape of recent price action, not about whether the stock is worth buying or selling.

The popular shorthand — high RSI means "due for a pullback," low RSI means "due for a bounce" — is a real pattern some of the time and a trap the rest of the time, and the difference is the single most important thing to understand about this tool:

A stock in a strong, sustained trend can sit at an extreme RSI reading for a long stretch — that's not the indicator malfunctioning, it's the indicator doing exactly what it's built to do, and reading "extreme therefore reverse" into it is where most of the damage happens.

Does a high or low RSI predict a reversal?

Not reliably by itself, and this is worth stating plainly rather than hedging around it. RSI extremes are a tendency, not a signal — they describe conditions that have often preceded a pause or a pullback historically, in ranging or moderately trending conditions. In a strong trend, "extreme" can persist far longer than the intuitive read expects, and traders who fade every overbought reading in a genuine uptrend tend to lose money to the trend, repeatedly, while waiting for a reversal that isn't due yet.

The honest use of RSI extremes is as one input that raises or lowers how much weight to put on other evidence — not as a standalone trigger. Overbought inside a choppy, range-bound stock is a very different piece of information than overbought inside a stock that just broke into a strong move: the first has real reversal odds behind it; the second is often just what a healthy trend looks like on this particular ruler.

What is RSI divergence, and is it more reliable than the raw reading?

Divergence is when price makes a new high (or low) but RSI does not confirm it with a matching new high (or low) of its own — price pushing further while the underlying gain/loss mix is quietly getting weaker. It's a more specific piece of information than a raw overbought/oversold reading because it's comparing the trend of momentum against the trend of price, not just reading one snapshot.

Divergence is more informative than a bare extreme reading, but it is still not a trade signal on its own — divergences can run for a long time before price actually turns, and plenty resolve with price catching back up to what momentum was already showing rather than momentum catching down to price. It belongs in the same category as everything else here: a piece of evidence, weighed alongside structure, not a trigger pulled in isolation.

How does RSI relate to trend strength, not just reversal timing?

A less-discussed but genuinely useful read: RSI tends to hold a higher range during an uptrend and a lower range during a downtrend, and watching which range it's oscillating in — rather than watching for a single extreme touch — says something about which regime a stock is actually in. A stock whose pullbacks keep finding support in the middle of the range, never dropping into oversold territory, is behaving differently than one that keeps round-tripping the whole scale. That's a regime read, closer in spirit to reading market breadth than to picking a single reversal point.

What does RSI get used wrong for?

• Treating an extreme reading as a standalone entry or exit signal. The number describes momentum, not a level to trade off of — the same mistake as treating ATR as a support/resistance level: a bounded oscillator reading is not a price, and it doesn't behave like one.
• Fading every overbought signal in a strong trend. This is the single costliest habit tied to this indicator — a trend that keeps producing overbought readings is often telling you it's strong, not that it's due. Selling into strength on RSI alone, repeatedly, in a real trend is a slow way to lose to a move that was never going to reverse on schedule.
• Ignoring the lookback period's effect. A 14-period RSI and a 5-period RSI on the same chart will disagree constantly — the shorter window swings to extremes far more often and means something different each time it does. Comparing readings across different periods as if they're the same measurement is a quiet source of confusion.
• Using RSI in isolation from everything else on the chart. It's one read of one thing — recent gain/loss lopsidedness — not a complete trading system by itself, any more than RVOL alone tells you the whole story of a session. MACD measures something different — the relationship between two moving averages, not a bounded gain/loss ratio — which is why traders often watch both rather than picking one.

Questions traders ask

What counts as overbought or oversold on RSI?

The traditional convention treats the upper end of the scale as overbought and the lower end as oversold, with the exact threshold traders use varying by stock, timeframe, and how the stock has behaved historically — a threshold tuned to one ticker's normal range is more useful than a fixed number applied identically to every chart, especially since a strong trend can sit past the conventional threshold for an extended stretch without reversing.

Is a shorter or longer RSI period better?

Neither is universally better — a shorter lookback reacts faster and swings to extremes more often, which suits traders looking for frequent, quick reads; a longer lookback smooths more and reacts slower, which suits a read on the larger trend. The 14-period default is a well-tested starting point, not a rule, the same way ATR's 14-period default is a convention rather than a law.

Can RSI be used the same way on every stock?

The construction is the same on every stock — it's a normalized, bounded scale by design — but how a specific stock's RSI has historically behaved (how often it reaches extremes, how long it stays there) differs by name, so a reading that's unusual for one stock can be perfectly normal for another. Comparing a stock's current RSI to its own history is a more useful habit than comparing raw readings across different tickers.

Does RSI work better in trending or ranging markets?

The overbought/oversold reversal read tends to work better in ranging or moderately trending conditions, where extremes really do mark exhaustion more often than not. In a strong, sustained trend, RSI can sit at an extreme for a long stretch without reversing, which is exactly the condition that makes the popular "fade every extreme" habit expensive — recognizing which regime a stock is in matters more than the raw reading itself.

RSI is a bounded momentum ruler: it measures how lopsided recent gains and losses have been, on a scale with a real floor and ceiling, which is what makes "overbought" and "oversold" mean anything at all. What it does not do is tell you whether that lopsidedness is about to end — that's a separate question, and on this desk it's answered by where a market actually sits in its larger cyclic structure, covered on the methodology page.

Keep reading