Pineify logo
TradingView indicators

Stochastic RSI: formula, K/D and Pine Script

The RSI stochastic indicator measures where RSI sits within its own recent range. Compute RSI first, normalize that series, then smooth the result into K and D. It is different from applying Stochastic to price. Use the worked table and copyable code to check each stage, including scale, missing values, and startup bars.

Example defaults: RSI 14 · Stochastic 14 · K 3 · D 3 · Close source

Inspect the Pine ScriptBuild a custom version

Two windows, followed by two smoothing steps

TradingView defines Stoch RSI as RSI relative to its own high-low range. RSI length controls the upstream momentum calculation. Stochastic length controls the rolling minimum and maximum of that RSI series.

R = RSI(source, RSI length)
L = lowest R over N bars; H = highest R over N bars
raw = (R − L) / (H − L)
K = SMA(raw, K length)
D = SMA(K, D length)

The stochastic window includes the current RSI reading. Multiplying raw, K, and D by 100 changes only their display scale. The reference levels 0.2 / 0.8 become 20 / 80; a comparison against 80 on a 0-1 series cannot work.

Normalization and smoothing do different jobs. A small RSI change can occupy a large fraction of a narrow RSI range. Averaging then spreads that movement across several observations. More smoothing does not make the response faster.

Worked example: RSI can fall while Stoch RSI rises

This hypothetical input is a sequence of already-computed RSI readings, not market data or a backtest: 40, 50, 60, 55, 45, 50, 55. Set the stochastic window to 3 and both smoothing lengths to 2. The shorter lengths make every calculation visible.

BarRSI3-bar low / highRaw × 100K × 100D × 100
140Incompletenanana
250Incompletenanana
36040 / 60100nana
45550 / 605075na
54545 / 6002550
65045 / 55502525
75545 / 551007550

At bar 5, raw = (45 − 45) / (60 − 45) = 0. K averages raw values 50 and 0 to give 25; D averages K values 75 and 25 to give 50. Raw reaches zero while both smoothed lines remain above zero.

Now consider a separate 3-bar window: RSI values 70, 40, 50 give raw = 33.33 on the 0-100 scale. Let the next RSI be 49. The new window 40, 50, 49 gives raw = 90. RSI fell from 50 to 49, yet raw rose because the old high of 70 left the window. Neither calculation predicts what price does next.

Pine Script with explicit missing-value rules

Paste this into TradingView Pine Editor and add it to the chart. It retains K/D plots and reference bands, adds raw values to the Data Window, and lets you switch scales. This is a calculation example, not a strategy.

Pine Script v6
//@version=6
indicator("Stochastic RSI - explicit windows", overlay = false, precision = 2, max_bars_back = 500)

rsiLength = input.int(14, "RSI length", minval = 1, maxval = 500)
stochLength = input.int(14, "Stochastic length", minval = 2, maxval = 500)
kLength = input.int(3, "K smoothing", minval = 1, maxval = 500)
dLength = input.int(3, "D smoothing", minval = 1, maxval = 500)
source = input.source(close, "RSI source")
unitScale = input.bool(false, "Use 0-1 scale")

// Require consecutive valid observations; do not skip missing bars.
strictAverage(float values, int length) =>
    float total = 0.0
    bool valid = true
    for i = 0 to length - 1
        if na(values[i])
            valid := false
        else
            total += values[i]
    valid ? total / length : na

rsi = ta.rsi(source, rsiLength)
var int validRsiBars = 0
validRsiBars := na(rsi) ? 0 : validRsiBars + 1
lowestRsi = ta.lowest(rsi, stochLength)
highestRsi = ta.highest(rsi, stochLength)
rsiRange = highestRsi - lowestRsi
ready = validRsiBars >= stochLength
raw = ready and rsiRange > 0 ? (rsi - lowestRsi) / rsiRange : na
kUnit = strictAverage(raw, kLength)
dUnit = strictAverage(kUnit, dLength)
scale = unitScale ? 1.0 : 100.0

plot(kUnit * scale, "K", color.blue, 2, style = plot.style_linebr)
plot(dUnit * scale, "D", color.orange, 2, style = plot.style_linebr)
plot(raw * scale, "Raw Stoch RSI", display = display.data_window)
upper = hline(unitScale ? 0.8 : 80.0, "Upper reference", color.gray)
hline(unitScale ? 0.5 : 50.0, "Midpoint", color.new(color.gray, 50))
lower = hline(unitScale ? 0.2 : 20.0, "Lower reference", color.gray)
fill(upper, lower, color.new(color.blue, 92))

Blue is K; orange is D. The shaded band runs from 20 to 80, or 0.2 to 0.8. Gaps are intentional: no value is drawn while the RSI range is zero or a required window is incomplete. This explicit policy can differ from an implementation that uses partial windows or skips missing values.

Parameters and the first valid bar

InputDefaultWhat changes
RSI length14The upstream RSI calculation, before normalization.
Stochastic length14Which RSI extrema define the current range. Minimum 2 in this example.
K smoothing3Consecutive raw values averaged into K. Set 1 to inspect raw.
D smoothing3Consecutive K values averaged into D. Set 1 to make D equal K.
RSI sourceCloseThe series supplied to ta.rsi. Keep it identical when comparing scripts.
Use 0-1 scaleFalseSwitches all displayed values and reference levels together.

Count from the first valid RSI observation. With N = 14, raw first becomes eligible on RSI observation 14, K on observation 16, and D on observation 18, provided the RSI range remains positive. The first valid RSI itself needs earlier price history; 18 RSI observations do not mean 18 price bars from an empty chart.

The general D requirement is N + K length + D length − 2 valid RSI observations when there are no gaps. The code requires consecutive valid raw and K windows. A zero-range raw value invalidates K until that gap leaves its averaging window, and the K gap propagates into D.

If the RSI window is [50, 50, 50], H − L is zero. The code returns na, not zero or the midpoint. With RSI [49.9, 50, 50.1], the raw reading is 100 despite a total RSI spread of only 0.2. A relative extreme says little about the absolute size of a price move.

How to read Stochastic RSI without turning a level into a forecast

ObservationWhat the calculation establishesWhat still needs testing
Raw = 100Current RSI equals its window high when the range is positive.Whether price continues or reverses.
K crosses DThe average of raw readings changes position relative to its own average.Any entry, exit, or confirmation rule.
K stays above 80Recent normalized RSI readings remain near their range highs.Whether a short position has any advantage.
A live cross disappearsThe current candle changed before it closed.Use a defined bar-close decision rule when testing.

Pine indicators recalculate on realtime updates. The code displays the live candle. Compare closed bars first; if you later add alerts, specify whether they require bar confirmation.

RSI, price Stochastic, and Stoch RSI share bounded displays but use different inputs. RSI processes gains and losses; Stochastic compares price to its price range. Stoch RSI applies range normalization to RSI. Matching 14-bar settings does not make these formulas interchangeable.

Educational calculation only, not investment advice. A strategy needs separate rules for order timing, exits, position sizing, costs, and missing observations. No profit or signal-accuracy claim is made here.

Turn the calculation into a specification for Pineify

The Pineify Pine Script AI Coding Agent can help write a custom script from a description. Give it the calculation contract before asking for alerts or strategy orders.

  1. Set the chart source and RSI/stochastic/K/D lengths explicitly.
  2. Specify the output scale and the zero-range policy: return na and require consecutive valid smoothing windows.
  3. Compare raw, K, and D against the table and inspect the first valid observations.
  4. Define a closed-bar event, such as K crossing D, separately from an order rule.
  5. Add exits, sizing, commission, and slippage before evaluating a strategy; keep a separate period for evaluation.
Open Pine Script AI Coding Agent

Stochastic RSI questions

Stochastic RSI measures RSI within its own recent range. It first computes RSI from a price source, then normalizes that RSI series. The result describes a relative position, not a probability of a price reversal.

In this example, K averages consecutive raw Stoch RSI values; D averages consecutive K values. A K/D cross describes their relative movement. A high smoothed K value does not mean the current RSI itself is exactly at the top of its range.

These are the same normalized values with a scale factor of 100. A value of 0.2 equals 20 and 0.8 equals 80. Change thresholds together with the scale. The code provides a switch that changes plots and reference levels together.

The raw formula divides by zero and has no defined range position. This implementation returns na, leaves a gap, and waits for complete valid smoothing windows before plotting K and D again. Returning 0 or 50 instead would be a different rule.

The code starts with RSI 14, stochastic length 14, K 3, and D 3. Treat this as a reproducible baseline. Change one input at a time and compare bar values before evaluating any trading rules; there is no universal best timeframe or setting.

It waits for a complete stochastic window of valid RSI readings and complete consecutive K/D windows. Other implementations may accept partial startup windows or skip missing observations during smoothing. Compare startup and missing-value policies before treating a difference as an error.

Paste the example into Pine Editor and add it to the chart. Match its source, chart timeframe, four lengths, and scale to the indicator you are comparing. Inspect raw, K, and D in the Data Window on closed bars; this script is an indicator and does not place strategy orders.

No. A cross is an oscillator event. A testable trading system still needs entry timing, exits, position sizing, commissions, slippage, and a rule for missing values. Intrabar crosses can disappear before the candle closes.

Sources and related indicators

Official references checked October 17, 2026. The worked RSI sequences and strict window policy are examples created for this page.