Momentum & oscillators

Relative Strength Index · RSI

A 0–100 gauge of how one-sided recent gains have been — the classic overbought / oversold oscillator.

Works best in ranging marketsEngine-computed on a fixed sample series
145120963070Above 70 = overboughtBelow 30 = oversold
RSI 87.65How to read RSI on the chart — the callouts mark what to look for.

The formula

Average the up-closes and the down-closes over the lookback using Wilder's smoothing, divide to get the relative strength RS, then map it onto a bounded 0–100 scale. 14 is standard; shorter (2–9) is used for mean reversion, longer (21+) to smooth.

RSI = 100 − 100 ÷ (1 + RS), RS = Average Gain ÷ Average Loss
Worked example
Avg gainAvg lossRSRSI
1.200.801.5060

RS = 1.20 ÷ 0.80 = 1.50 → RSI = 100 − 100 ÷ 2.5 = 60

What the RSI is and the question it answers

The Relative Strength Index, published by J. Welles Wilder in his 1978 book New Concepts in Technical Trading Systems, is the most widely used momentum oscillator in all of technical analysis. It condenses the tug-of-war between buyers and sellers into a single line that travels between 0 and 100, answering one question: over the recent lookback, how one-sided have the gains been relative to the losses? A high reading means up-closes have overwhelmingly dominated and the advance may be stretched; a low reading means down-closes have dominated and the decline may be overdone. Despite the name, it has nothing to do with comparing one asset's strength against another — the strength is internal, price against its own recent history. For a complete beginner, think of it as a speedometer for momentum that also flashes a warning when the move looks unsustainably fast in either direction.

How the RSI is calculated

For each bar, take the change from the prior close; a positive change is a gain and a negative one is a loss, recorded as a positive number. Wilder then smooths these separately into an average gain and an average loss over the lookback, defaulting to 14 periods. The first values are simple averages of the first 14 gains and losses; thereafter each new average is the prior average times 13, plus the current value, all divided by 14 — Wilder's smoothing, which is a slow exponential average. Relative strength is the average gain divided by the average loss, and the RSI is 100 minus 100 divided by one plus that relative strength, which mathematically forces the result between 0 and 100. When average loss is zero the formula returns 100, and when average gain is zero it returns 0.

Reading the RSI, step by step

The textbook levels are 70 for overbought and 30 for oversold, with 50 as the centerline that separates net-bullish from net-bearish momentum. But those lines are context, not triggers — in a strong uptrend RSI can hold above 70 for a long stretch, and shorting simply because it is overbought is the single most common way traders lose money with the tool. The richer signal is divergence: price posts a higher high while RSI posts a lower high (bearish), or price a lower low while RSI a higher low (bullish), hinting the move is running out of fuel. Constance Brown and Andrew Cardwell refined this further — in a bull market RSI tends to oscillate between roughly 40 and 80, using 40 to 50 as support, while in a bear market it lives between 20 and 60, using 50 to 60 as resistance. Failure swings, where RSI breaks a prior oscillator low or high without a corresponding price break, are Wilder's own confirmation of a turn.

Reading the signals on the chart

145120963070
RSI 87.65The ▲/▼ marks flag the most recent crossings of the 30 and 70 lines — the classic oversold / overbought signals.

Best timeframes

  • Scalping1m – 5mperiod 2–7, fast
  • Day trading5m – 15mperiod 14 standard
  • Swing1H – daily
  • PositionDaily – weeklyperiod 14–21

The 14-period RSI is the reference on every timeframe; drop the length for mean reversion, raise it to smooth.

RSI vs other momentum tools

RSIStochasticMACD
Bounded 0–100YesYesNo
MeasuresGain vs lossClose in rangeEMA gap
Best signalDivergence%K/%D crossSignal cross

Common price-action setups

How the signal typically plays out on the chart.

Oversold reclaim

In a range, RSI turns up from below 30 and reclaims it — buy the reclaim with a stop under the swing low that printed the oversold reading.

Buy 30 reclaim
Mean-reversion bounce
Overbought rejection

RSI rolls back under 70 in a range — sell or trim longs, placing the stop above the swing high that made the overbought print.

Sell 70 rollover
Pullback lower
Bullish divergence

Price makes a lower low while RSI makes a higher low — enter as RSI turns up, risking just below the price low. Confirm with structure before committing.

Buy divergence
Trend reversal up

Best timeframes and settings

RSI is remarkably timeframe-agnostic and is used everywhere from one-minute scalps to monthly charts, but its behavior changes with the lookback. The default 14 is a sensible all-purpose swing setting on daily and 4-hour charts. Short-term and mean-reversion traders often drop to a 2- to 9-period RSI, which reaches the extremes quickly and suits Larry Connors-style oversold-bounce systems; longer settings of 21 or more smooth the line for position trading and cleaner divergences. The trade-off is the familiar one: a shorter period is more responsive and generates more signals but far more noise and false extremes, while a longer period is calmer and more reliable but slower to react. Match the period to your holding time — a 2-period RSI on a monthly chart is nonsense, as is a 21-period on a one-minute scalp.

When and where to use it

RSI earns its keep in ranging and mean-reverting markets, where fading turns out of overbought and oversold has a genuine edge. In trending markets its role flips from a reversal tool to a pullback tool: buy dips when RSI eases to the 40 to 50 support band in an uptrend rather than betting on a reversal from 70. It is applied across every liquid asset class — equities, indices, futures, FX, and crypto — because it only needs a clean close series. The regime to avoid is a strong, persistent trend traded as if it were a range; that is where overbought and oversold readings become bull and bear traps. On illiquid or heavily gapping instruments the single-bar change that feeds RSI becomes distorted, so treat those readings with caution.

Strategies that use the RSI

The classic range strategy fades extremes — go long as RSI turns up through 30 from below and short as it rolls down through 70 from above, with stops beyond the recent swing and targets at the range midline or opposite band. The trend-pullback strategy, better suited to trends, waits for an established uptrend and buys when RSI pulls back to 40 to 50 and turns up, stopping under the swing low and trailing with the trend. The divergence-plus-trigger strategy is the most disciplined: identify a bearish divergence at a price high, then wait for a break of the last minor swing low or a 50-line cross before entering short, which filters out the many divergences that never resolve. A fourth, the Connors 2-period system, buys a stock above its 200-day average when a 2-period RSI drops below 5 to 10 and exits on a move back above 50 or a close above a short moving average.

Combining the RSI with other indicators

RSI is strongest when a trend filter tells you which mode to use — a 200-day SMA or the MACD zero line distinguishes a range you can fade from a trend you should only buy on dips. Horizontal support and resistance or Fibonacci levels give an overbought or oversold turn a price reason to reverse, dramatically improving reliability over trading the number alone. Candlestick reversal patterns at an RSI extreme — a hammer at oversold, a shooting star at overbought — provide the bar-by-bar confirmation Wilder's failure swings were reaching for. Volume tools such as OBV help validate whether a divergence has real distribution or accumulation behind it. Pairing RSI with a differently built oscillator like MACD, which measures trend momentum rather than bounded overbought and oversold, gives independent confirmation without simple duplication.

Where the RSI fails

The defining failure is the overbought-equals-sell reflex; a strong uptrend can keep RSI above 70 for weeks, turning every premature short into a loss. Divergence can persist across many bars — a market can be overbought and diverging and still climb another 20 percent — so divergence without a confirming trigger is an invitation to be early. In quiet, choppy tape RSI crosses 50 back and forth with no follow-through, generating whipsaw after whipsaw. Very short lookbacks amplify all of these problems, throwing false extremes on ordinary noise. The universal remedy is to demand confirmation from trend, structure, or price action and to let the market prove the turn before committing, rather than front-running the oscillator.

A worked example

Take a daily chart with the default 14-period RSI. After a run-up, the average gain over the window is 1.2 points and the average loss is 0.4 points, so relative strength is 3.0 and RSI is 100 minus 100 divided by 4, which equals 75 — clearly overbought. A novice shorts immediately; the trend continues and RSI stays above 70 for two more weeks, stopping them out. The disciplined trader instead notes that price soon makes a higher high at 152 while RSI makes a lower high at 68 — a bearish divergence — and waits. When price then breaks the last minor swing low at 148, the momentum failure is confirmed; they short at 147.50 with a stop at 152.20 above the divergent high, targeting the 50-day average near 138. The RSI reading flagged the risk, but the trade only triggered once price confirmed the turn.

Common mistakes

  • Shorting a strong uptrend just because RSI is overbought — the single most common way traders lose money with it.
  • Trading every 30 / 70 touch with no regard for the trend or structure.
  • Expecting divergence to resolve immediately; it can persist for many bars.
  • Using the default 14 for mean reversion without testing a shorter length.
  • Forgetting that in a trend RSI can pin in the extreme zone for a long time.
  • Treating the 70 / 30 lines as signals rather than context.