Guide · Indicators
The RSI indicator, explained: how it is built, how to read it, and where it misleads
The short answer
The Relative Strength Index (RSI) is a momentum oscillator, published by J. Welles Wilder in 1978, that scores recent price action on a bounded 0 to 100 scale. It is 100 − 100 / (1 + RS), where RS = average gain / average loss over a lookback of N periods, default 14, smoothed the way Wilder specified. Despite the name it does not compare one stock to another; it measures the internal ratio of up-moves to down-moves in a single instrument. It is a bounded derivative of price, so it summarises momentum that has already happened and cannot lead the price it is built from. An oscillator describes momentum. It does not schedule reversals.
RSI is the most-used momentum indicator on Indian retail platforms, and the most misread. Almost every explanation stops at "above 70 is overbought, below 30 is oversold, so sell high and buy low." The mechanism says otherwise. Overbought is a statement about momentum, not a sell instruction, and in a strong trend the reading can stay stretched for weeks while price keeps going. Rather than assert that, this page lets you watch it: the interactive visualizer below computes RSI on a generated price series and shows, live, how the line stays pinned in the band during a trend and oscillates in a range. Then it builds the formula from first principles, reads the bands by regime, and spends real time on the failure mode that costs people the most.
Interactive · see it, do not just read it
RSI regime visualizer
Choose a market regime and an RSI lookback, and watch what the oscillator actually does. The price series is generated and deterministic, drawn purely to teach the mechanism, it is not market data. The one thing to notice: in a trend, RSI reaches the overbought or oversold band and stays there while price keeps running. That is the whole point. Overbought is not a sell.
14
Bars in the band
−
Longest unbroken stretch
−
Final RSI reading
−
Times it crossed the band
−
Pick a regime to begin.
Deterministic illustrative series (fixed seed per regime), computed with Wilder's smoothing exactly as the formula section below describes. No market data, no signals, no claim about any real instrument. Change the period to 2 and then to 28 to feel how sensitivity trades against whipsaw.
What just happened, and why it matters
If you left the visualizer on strong uptrend, RSI vaulted into the overbought band early and sat there, sometimes for the whole advance, while price kept making new highs. Every crossing of 70 that a mechanical system would have sold was followed by higher prices. Switch to range and the behaviour inverts: RSI swings from one band to the other, and the extremes actually do mark the turns. That single contrast is the most important thing on this page, and it is why "overbought means sell" is folklore rather than mechanism.
The rest of this guide earns that observation. It builds the formula so the readings stop feeling like magic, works one update by hand, sets out how the bands read differently by regime, and then confronts the edge cases where even a correct-looking RSI misleads. The SEBI FY25 finding that over 91 percent of individual F&O traders were net loss-making is, in part, an indicator-worship story: a stretched oscillator was read as an instruction, and the instruction was to fade a market that was trending. The visualizer is the antidote, watched rather than argued.
What RSI is, and the name that misleads
Start with the word that trips everyone. "Relative Strength Index" sounds like it ranks one security against another, the way a sector leader is said to have strength relative to the index. It does not. RSI compares an instrument against its own recent history: how large the up-closes have been versus the down-closes over the lookback. That is a measure of internal momentum, the velocity and persistence of a single price series, not a cross-sectional comparison. Wilder chose the name in 1978, and the ambiguity has confused readers ever since. Keep the correct definition fixed: RSI is a one-instrument momentum gauge, bounded between 0 and 100.
Because it is bounded, RSI behaves as an oscillator. Values cannot run away to infinity the way a raw price or an unbounded indicator can; they are compressed into a fixed band. That is the whole point of the final step in the formula. It takes an open-ended ratio, average gain over average loss, and squeezes it into a range you can read at a glance and compare across instruments and time. A reading of 65 on an index means the same structural thing as 65 on a large-cap: up-moves have moderately outweighed down-moves lately. The bounding is what makes 70 and 30 usable as fixed reference lines at all. It is also the source of a limit worth stating early: because RSI is computed from price, it is a derivative of price and can never lead it. It reports; it does not predict.
How RSI is built: the formula, step by step
The construction is short, and worth doing by hand once so the readings stop feeling like magic. Over the lookback of N bars (default 14), you separate each bar's close-to-close change into an up-move or a down-move. Losses are recorded as positive numbers. Then you average each side, take their ratio, and bound it.
Two details separate a correct RSI from a rough one. First, Wilder's smoothing. The very first average gain and loss are plain N-period averages, but every bar afterward is a running average: the new average equals the previous average multiplied by (N − 1), plus the current bar's value, divided by N. This is a smoothed, exponential-style average with a smoothing factor of 1 over N, not a fresh simple mean each bar, which is why an RSI computed with simple moving averages will not match the standard line on charting software. The visualizer above uses exactly this recurrence. Second, the bounding is exact: if there have been no losses in the window, RS is undefined and RSI is defined as 100; if there have been no gains, RSI is 0. The oscillator can touch its rails, and near them the readings become jumpy because tiny denominators swing RS violently.
A worked calculation you can follow
Numbers make the smoothing concrete. Suppose over the first 14 bars the up-closes averaged 1.00 point and the down-closes averaged 0.50 point (losses recorded as positive). Then RS is 1.00 / 0.50 = 2.0, and RSI is 100 − 100 / (1 + 2.0) = 100 − 33.3 = 66.7. Now the fifteenth bar closes up by 1.30. Wilder's smoothing updates each side before recomputing.
| Step | Figure | How it is derived |
|---|---|---|
| Prior average gain | 1.00 | Mean of the up-closes over the first 14 bars |
| Prior average loss | 0.50 | Mean of the down-closes over the first 14 bars |
| New average gain | 1.02 | (1.00 × 13 + 1.30) / 14, the new bar was an up-move |
| New average loss | 0.46 | (0.50 × 13 + 0) / 14, no loss on this bar |
| New RS | ≈ 2.21 | 1.02 divided by 0.46 |
| New RSI | ≈ 68.9 | 100 − 100 / (1 + 2.21) |
The reading rose from 66.7 to about 68.9 on a single up-bar, a small step, not a jump, because each bar is only one-fourteenth of the smoothed average. That inertia is a feature: it stops one loud candle from throwing the indicator around. It is also why RSI is a lagging summary, it weights the whole recent window, so by construction it describes momentum that is already in the tape. The math is illustrative, chosen for clean arithmetic, not drawn from any real instrument.
How to read it: bands, the midline, failure swings, divergence
RSI offers four readings, and they are not equally strong. The weakest, and the most quoted, is the threshold cross. The strongest are the ones that use the shape of the line rather than a single level.
| Reading | What it is | What it reflects | The caveat |
|---|---|---|---|
| 70 / 30 bands | Fixed overbought and oversold lines (80 / 20 for a stricter variant) | Up-moves have strongly out- or under-weighed down-moves lately | Only marks turns in a range; in a trend the line stays pinned |
| 50 midline | The RS = 1 balance point | Above 50, buyers have had the edge over the window; below, sellers | A bias gauge, not a trigger; it whipsaws in a flat, choppy market |
| Failure swing | A high or low the line fails to exceed, then breaks the interim level | Momentum stalled and rolled over inside the oscillator itself | Independent of price, but still only a hint of a possible turn |
| Divergence | Price and RSI make opposite highs or lows | Price extended further on less momentum than before | Can persist and fail repeatedly in a strong trend; not a timing signal |
The 50 midline is underused and quietly the most robust of the four. Because RSI sits at 50 exactly when average gain equals average loss, staying above 50 says buyers have held the balance over the lookback and staying below says sellers have. Read as a bias filter rather than a signal, the midline often carries more information than the 70 and 30 extremes, precisely because it does not assume the market must revert. It is the reading the visualizer's trend regimes respect: in the uptrend the line rarely loses 50, even as it holds above 70.
Reading RSI by regime: the same number, three meanings
A single RSI value has no fixed meaning until you know the regime it sits in. This is the interpretation table competitors skip, because it refuses to hand out a rule. The same 75 that is a warning in a range is confirmation of strength in a trend. Read down the column that matches the market you are actually in, and notice that the honest read in a trend is almost never "fade the band."
| RSI reading | In a strong uptrend | In a range | In a strong downtrend |
|---|---|---|---|
| Above 70 | Momentum, not a top: trend confirmed, the line can stay here for weeks | Upper edge of the range: mean-reversion pressure often builds | Rare counter-trend bounce: usually the weakest part of the down move |
| Around 50 | Healthy pullback within the trend, buyers still in control over the window | The middle of the range: least informative reading there is | A bounce that has not reclaimed control; sellers still hold the window |
| Below 30 | Sharp, shallow dip: often a pause in an uptrend, not a reversal | Lower edge of the range: mean-reversion pressure often builds | Weakness, not a bottom: the line can stay pinned here as price falls |
| Holding one side of 50 | Above 50 persistently: the cleanest read that the trend is intact | Crosses 50 often: the whipsaw that makes the midline noisy in a range | Below 50 persistently: the cleanest read that the down-trend is intact |
The table is the visualizer written as prose. When you set the tool to an uptrend, "above 70" lit up for most of the series and price rose the whole time, which is the top-left cell. When you set it to a range, the same band marked the edges and price snapped back, which is the middle column. The number did not change meaning; the regime around it did. This is why a serious reading of RSI begins by identifying the regime and only then consults the oscillator, never the other way round.
The failure mode that costs the most: overbought is not sell
Here is the single most important thing to understand about RSI, and the part most explanations bury. Overbought does not mean sell. It means momentum has been strong, and strong, persistent momentum is the definition of a trend. In a powerful uptrend, RSI can climb above 70 and simply stay there, week after week, while price keeps making new highs, a behaviour StockCharts documents explicitly: momentum oscillators can become overbought and remain so in a strong up-trend, and oversold and remain so in a strong down-trend. Every mechanical short taken because "RSI is overbought" is a short into a rising market, and the market runs it over. The mirror is just as brutal on the downside: in a hard decline RSI can sit below 30 for a long time while price keeps falling, so "oversold" buyers catch a knife again and again.
The resolution is not a better threshold; it is knowing which regime you are in. Overbought and oversold are meaningful only in a ranging market, where price oscillates around a mean and extremes tend to snap back. In a trending market the same extremes are the wrong tool: the correct read there is the midline and the direction of momentum, not a fade of the band. This is the crux of the matter, so it is worth cross-reading the dedicated piece on the myth that oversold means buy, which walks through why "oversold" in a downtrend is a trap rather than a bargain. Divergence carries the same warning in a subtler form: it can persist for a long time without a reversal, so a divergence is a note that momentum is thinning, never a countdown to a top. If you want to see how professionals separate the divergences worth noting from the ones that keep failing, the deep dive on regular versus hidden divergence takes it further.
| Market regime | What price does | Does fading 70 / 30 make sense? | The better RSI read |
|---|---|---|---|
| Ranging | Oscillates around a mean between support and resistance | Sometimes: extremes often revert here | Bands can mark the edges of the range; still only a hint |
| Trending | Makes sustained higher highs or lower lows | No: the band stays pinned while price runs | The 50 midline and momentum direction, not the extreme |
| Transition | Shifts from range to trend or back | Unreliable: the regime itself is unclear | Wait for structure to resolve; RSI alone cannot tell you |
Period and timeframe: the sensitivity tradeoff, with numbers
The default is RSI(14) on the daily chart, and it is the default for a reason: Wilder specified 14, and it remains the near-universal setting across charting software and across Indian indices such as Nifty 50 and Bank Nifty. A common instinct is to shorten the lookback to make RSI "more responsive." It does become more responsive, and that is precisely the problem. Every reduction in N buys sensitivity at the direct cost of smoothness, and past a point that is not more information, it is more noise. The table sets the tradeoff out explicitly, and the visualizer lets you feel it: drag the period to 2 and the line slams between the rails; drag it to 28 and it barely leaves the middle.
| Period N | Latest-bar weight (1/N) | Sensitivity | How often it touches 70 / 30 | Main failure mode |
|---|---|---|---|---|
| 2 to 5 | 50% to 20% | Very high | Constantly; reaches the rails intraday | Whipsaw: extremes so frequent they mean little |
| 7 to 9 | 14% to 11% | High | Often | More signals, a higher share of them false |
| 14 (default) | ≈ 7% | Balanced | Occasionally, at genuine extremes | Lag: it confirms momentum already in the tape |
| 21 to 30 | 5% to 3% | Low | Rarely; mostly sits nearer 50 | Sluggish: turns are well underway before it reacts |
The same arithmetic explains the timeframe. RSI(14) does not care whether a bar is a day or five minutes; it looks back 14 bars either way. On a 5-minute chart that is only 70 minutes of memory, so the line touches the bands many times a session, and each touch carries the same weak information that a short daily period does. Dropping the timeframe is, in effect, shortening the lookback in wall-clock terms. Neither move reveals more truth; both trade smoothness for sensitivity, and in doing so usually manufacture whipsaw rather than insight. There is no "best" period, only a period matched to how much noise you are willing to read as signal.
Edge cases: where a correct-looking RSI still misleads
A right formula does not make a reading safe. Five conditions detach what RSI appears to say from what is actually happening, and each one has cost retail traders who were, on paper, reading the indicator correctly.
1. It is a derivative of price and cannot lead. RSI is computed entirely from closes, so it contains no information price does not already contain. It can only rearrange the recent past into a bounded number. Any use of RSI that treats it as a forecast, a reading that "predicts" the next move, is asking a mirror to see round a corner. It describes momentum that has happened; the future is not in the window.
2. Divergence that persists and keeps failing. In a strong trend, momentum can thin for a long stretch while price grinds on, printing bearish divergence after bearish divergence with no top. A trader who shorts each one is fighting the trend repeatedly on a signal that is, in that regime, close to noise. Divergence is a note about momentum, not a countdown; in a powerful trend it is the least reliable of the four readings, not the most.
3. Daily versus intraday are different instruments. The same RSI(14) is smooth on the daily and frantic on the 5-minute, because 14 bars is 14 days versus 70 minutes. A level that marks a real extreme on the daily chart is reached and abandoned dozens of times intraday. Reading an intraday RSI with daily-chart intuitions, or the reverse, imports the wrong sense of how meaningful a touch of the band is.
4. An index reads differently from a single stock. An index such as Nifty 50 is an average of many names, so single-stock shocks are diluted and its RSI trends more cleanly, which is exactly where fading an overbought reading is most dangerous. A single stock can gap on stock-specific news and throw RSI to an extreme in one bar. The identical 72 means "durable trend" more often on the index and "one-off event" more often on the stock. The number is comparable; the thing it summarises is not.
5. The 70 and 30 bands are conventions, not laws. Wilder proposed them; they are not derived from any market constant. In a quietly trending instrument 80 and 20 are often the more honest thresholds, and in a violent one even those are reached routinely. Treating 70 as a hard line where "overbought" begins is mistaking a 1978 convenience for a property of price. The band is a reading aid you set, not a boundary the market respects.
RSI myths versus what the mechanism actually says
Most RSI errors are the same handful of misreadings repeated. Setting the folklore beside the mechanism is the fastest way to inoculate against them.
| The myth | What the mechanism says |
|---|---|
| "RSI above 70 means sell" | It means momentum is strong. In a trend that can persist for weeks while price keeps rising. Overbought is not a signal to sell. |
| "RSI compares this stock to the market" | No. It compares an instrument to its own recent up-moves and down-moves. It is internal momentum, not a cross-security ranking. |
| "A shorter period is a better, faster RSI" | It is only more sensitive. It reaches extremes more often and generates more false readings. Sensitivity is not accuracy. |
| "Divergence means a reversal is coming" | Divergence flags fading momentum. It can persist and fail repeatedly in a strong trend. It is a hint, not a timer. |
| "RSI predicts the next move" | It is a bounded derivative of price. It summarises momentum already in the tape; it cannot lead the price it is built from. |
The one-line rule. An oscillator describes momentum; it does not schedule reversals. RSI is a good summary of recent momentum and a poor oracle. Every reading on this page is descriptive; none of it is a recommendation to buy, sell, or hold anything.
The Indian context, and where the real work is
Indian index behaviour makes the trend caution especially relevant. Broad Indian indices can trend hard and persistently, so the "overbought must fall" reflex is dangerous exactly where retail traders reach for it most, and the SEBI FY25 study, over 91 percent of individual F&O traders net loss-making, is a base rate that indicator-worship feeds directly. A stretched oscillator gets read as an instruction, the instruction is to fade a trend, and the trend obliges by continuing. None of this makes RSI(14) a rule to trade mechanically. It is a well-chosen default for a summary statistic, and the honest use of it is as one lens on momentum, cross-checked against the actual structure of the market, never as a standalone trigger.
Reading the regime, deciding whether an extreme even means anything here, is the judgement that separates a trader from a person watching a coloured line. That upstream work, identifying the regime before consulting the oscillator, is exactly what the method we teach is built around, and it is the part no indicator can do for you. The visualizer on this page is a small demonstration of a larger principle: an indicator is only as good as the reader's grasp of the market it sits on. The indicator is the easy part.
Common Questions
Frequently Asked Questions
What does the RSI indicator actually measure?
+RSI measures momentum: the ratio of recent up-moves to recent down-moves in one instrument, expressed on a bounded 0 to 100 scale. Despite the name, it does not compare the relative strength of two securities. It compares a stock, or an index, against its own recent behaviour. A high reading means up-days have dominated down-days over the lookback; a low reading means the reverse. It summarises what price has already done, it does not forecast what it will do next. It is a bounded derivative of price, so it cannot lead price; it can only describe the momentum already in the tape.
What is the RSI formula?
+RSI equals 100 minus 100 divided by 1 plus RS, where RS is the average gain divided by the average loss over N periods, default 14. The first average gain and loss are simple averages of the up-closes and down-closes over the first N bars. Every bar after that uses Wilder's smoothing: the new average equals the previous average times N minus 1, plus the current value, all divided by N. The 100 minus 100 over 1 plus RS step compresses the open-ended ratio into a fixed 0 to 100 range, which is why 70 and 30 can serve as fixed reference lines at all.
What do the 70 and 30 levels on RSI mean?
+They are the conventional overbought and oversold bands Wilder suggested in 1978, not laws. Above 70, up-moves have strongly outweighed down-moves recently; below 30, the reverse. Some readers use 80 and 20 to demand a more stretched reading before flagging an extreme, and in a trending instrument that stricter band is often more honest. The key caution is that these bands describe momentum, not a signal. In a ranging market they mark likely turning zones; in a strong trend RSI can sit above 70, or below 30, for weeks while price keeps going. The visualizer on this page lets you watch exactly that happen.
Why is overbought not a sell signal?
+Because overbought only means momentum has been strong, and strong, persistent momentum is what a trend is made of. In a sustained uptrend RSI can remain pinned above 70 for weeks while price keeps rising, so mechanically selling every time it crosses 70 puts you short against a rising market and gets run over. StockCharts documents that RSI can stay overbought through a strong up-trend and oversold through a strong down-trend. Overbought and oversold are meaningful in a ranging market, where price oscillates around a mean. In a trend they are the wrong tool, and treating them as a reversal trigger is the single most common RSI mistake. An oscillator describes momentum; it does not schedule reversals.
What is RSI divergence, and can it fail?
+Divergence is a disagreement between price and RSI. Bearish divergence is price making a higher high while RSI makes a lower high: price is still rising, but each push has less momentum behind it. Bullish divergence is the mirror below. Wilder saw divergence as a strong hint of a possible turn, but it is a hint, not a timer, and it fails routinely in strong trends. A powerful uptrend can print several bearish divergences before any top forms, and a hard downtrend can print bullish divergences that keep failing. Divergence flags thinning momentum; it never counts down to a reversal.
What is a failure swing in RSI?
+A failure swing is a pattern entirely inside the RSI line that does not depend on price. A top failure swing is RSI pushing above 70, pulling back, then failing to make a higher high on the next attempt and breaking below the intervening low. A bottom failure swing is the mirror below 30. Wilder treated failure swings as one of the stronger reversal hints the indicator offers, because the loss of momentum is confirmed within the oscillator itself rather than inferred from a single threshold cross. It is still a hint about momentum, not a guarantee about price.
Is a shorter RSI period better, and how does it differ intraday?
+Not better, just noisier. RSI of 14 on the daily chart is the default across Indian indices such as Nifty 50 and Bank Nifty and across charting software. Shortening the lookback to 7 or lower makes the line react to every small move, so it reaches 70 and 30 far more often and produces more extreme, more frequent readings. That is more signals but a far higher share of false ones. The same effect appears when you drop to intraday timeframes: a 14-period RSI on a 5-minute chart looks back only 70 minutes, so it whips between extremes many times a day. A shorter period, or a shorter timeframe, trades smoothness for sensitivity; it does not reveal more truth. Try period 2 versus period 21 in the visualizer to see the difference directly.
Is RSI on an index different from RSI on a single stock?
+The formula is identical, but the character of the readings differs. An index such as Nifty 50 is an average of many constituents, so idiosyncratic single-stock jumps are diluted and its RSI tends to be smoother and to trend more cleanly, which is exactly the condition in which fading an overbought reading is most dangerous. A single stock can gap on stock-specific news, sending RSI to an extreme in one bar that no amount of smoothing anticipated. So the same 72 reading carries a different texture on an index than on one name: on the index it more often means a durable trend, on the stock it can mean a one-off event. The number is comparable; the thing it summarises is not.
Does Bharath Shiksha give RSI buy or sell signals?
+No. Bharath Shiksha is an educational publisher, not a SEBI-registered investment adviser or research analyst. This page explains how RSI is constructed, how its readings are conventionally interpreted, and where the interpretation breaks down. It does not provide live calls, best settings, buy or sell recommendations, or any claim about accuracy or returns. The visualizer generates a deterministic illustrative series purely to teach the mechanism; it is not market data. An indicator summarises past price on a chart; the judgement about what, if anything, that means is the reader's own.
Where the facts come from
Sources
- J. Welles Wilder, New Concepts in Technical Trading Systems (1978). The original text that introduced the Relative Strength Index, the 100 minus 100 over 1 plus RS formula, Wilder's smoothing (new average = prior average times N minus 1, plus current, over N), the default 14-period lookback, and the failure-swing and divergence readings.
- StockCharts ChartSchool, Relative Strength Index (RSI). Establishes the step-by-step calculation, the simple first average then Wilder's smoothed running average, and the documented behaviour that momentum oscillators can become overbought and stay so in a strong up-trend, and oversold and stay so in a strong down-trend, with numerous divergences failing before a trend ends. chartschool.stockcharts.com
- Relative strength index, Wikipedia. Corroborates that the name is a misnomer, RSI measures the internal strength of a single instrument rather than a comparison between two, states the formula and Wilder's smoothing, and summarises the divergence and failure-swing concepts as Wilder defined them. en.wikipedia.org
Educational note. This guide explains how a momentum indicator is built and read. The visualizer computes RSI on a deterministic, generated price series to illustrate the mechanism; it is not market data and describes no real instrument. Nothing here is a recommendation to trade or invest, it is not a set of buy or sell signals, there are no best-settings, accuracy, or return claims, and it is not investment advice. Bharath Shiksha is an educational publisher, not a SEBI-registered investment adviser or research analyst.