Pineify logo
TradingView indicator guides

Deviation-Scaled Moving Average on TradingView

Ehlers DSMA adjusts an EMA-style weight using a filtered price change divided by its recent scale. Its response depends on that ratio, so a more volatile market does not automatically produce a faster line. This guide separates the original RMS approach from the standard-deviation EDSMA variant and gives you a Pine example with explicit startup and closing-bar rules.

alpha = 5 × |filtered / scale| / period
DSMA = previous DSMA + alpha × (source − previous DSMA)

When alpha is 0.25, the line closes one quarter of the current gap. An uncapped alpha above 1 can overshoot the source.

Inspect the Pine Script example

Specify which DSMA you mean

The July 2018 Traders' Tips implementation attributed to John Ehlers uses a two-pole Super Smoother with period N/2 and an RMS denominator. The public TradingView script by everget also uses the DSMA name; its release notes mention a choice of filter poles. Names alone are insufficient for numerical comparison. Sources: published implementation and TradingView author page.

Algorithm differences that change the plotted line
ChoicePublished RMS DSMACurrent Pineify EDSMA helper
Price inputSource minus source two bars agoSame difference, with a zero fallback for missing history
Super SmootherTwo poles; filter period N/2Two poles; fixed filter period 20
ScaleRMS: square root of mean squared filter valuesPopulation standard deviation of filter values
Alpha5 × absolute normalized filter / NSame ratio form, different denominator
StartupPlatform history and seed matterPrior DSMA uses a zero fallback

RMS measures distance from zero; standard deviation measures distance from the window mean. For a hypothetical filter window [2, 2, 2, 2], RMS is 2 and standard deviation is 0. Thus the RMS normalized value is 1, while the zero-denominator fallback in the current EDSMA helper gives 0. A nominally zero-mean filter need not have a zero mean in every finite window.

The example below implements the RMS choice with its own stated initialization. It is an educational reference, not a replacement for Pineify's built-in helper or a promise to match every community script. At N = 40 the two filter periods coincide, but the denominators and startup rules still differ.

How changing volatility changes the response

These are independent hypothetical single-bar calculations, not a backtest. Hold N = 40, previous DSMA = 100, and source = 110. Prices and filter values use the same arbitrary price units. Each result is 100 + alpha × 10; the normalized ratio and alpha are unitless.

Identical price gap, different filter-to-scale ratios
Scenario|Filter|RMSRatioAlphaNew DSMA
Small normalized move240.50.0625100.625
Larger current move8420.25102.500
Both values double480.50.0625100.625
Scale rises alone280.250.03125100.3125

A scale increase can reduce the response, while an unusually large current filtered value can increase it. Repeated noise can still move the line and create crossovers. The filter and the final smoothing both retain history; adaptation does not remove lag or establish a profitable entry rule.

For the RMS variant with the current value inside an N-value window, |filter/RMS| cannot exceed √N. Hence raw alpha cannot exceed 5/√N. At N = 40 the bound is about 0.7906; at N = 4 it is 2.5. Short periods can therefore overshoot. A standard-deviation denominator has no equivalent bound when the window mean is nonzero. Enabling an alpha cap changes the algorithm and must be part of the version you record.

Pine Script v6 with explicit initialization

The two-pole recursion filters the average of the current and previous two-bar price differences. The code uses exact π and √2 rather than the rounded constants in the published listing. RMS is the square root of the N-bar mean of squared filter values. Copy this standalone indicator into TradingView's Pine Editor and add it to a chart.

Pine Script v6: RMS variant

//@version=6
indicator("DSMA: RMS reference with explicit warm-up", overlay=true)
// Formula reference: TASC Traders' Tips, July 2018, EhlersDSMA.
// This example uses exact mathematical constants and explicit startup rules.
source = input.source(close, "Source")
length = input.int(40, "RMS period", minval=2)
capAlpha = input.bool(false, "Cap alpha at 1 (modified response)")

angle = math.sqrt(2.0) * math.pi / (0.5 * length)
decay = math.exp(-angle)
c2 = 2.0 * decay * math.cos(angle)
c3 = -decay * decay
c1 = 1.0 - c2 - c3

difference = bar_index >= 2 ? source - source[2] : 0.0
var float filtered = 0.0
filtered := c1 * (difference + nz(difference[1])) / 2.0 + c2 * nz(filtered[1]) + c3 * nz(filtered[2])
rms = math.sqrt(ta.sma(filtered * filtered, length))
ready = bar_index >= length + 1 and not na(rms)
scaled = ready and rms > 0.0 ? filtered / rms : 0.0
rawAlpha = 5.0 * math.abs(scaled) / length
alpha = capAlpha ? math.min(rawAlpha, 1.0) : rawAlpha

var float dsma = na
dsma := ready ? alpha * source + (1.0 - alpha) * nz(dsma[1], source) : source
plot(ready ? dsma : na, "DSMA", color=color.blue, linewidth=2)
plot(ready ? alpha : na, "Alpha", display=display.data_window)
plot(ready ? rms : na, "Filter RMS", display=display.data_window)

crossUp = ta.crossover(source, dsma)
crossDown = ta.crossunder(source, dsma)
confirmed = ready and ready[1] and barstate.isconfirmed
alertcondition(confirmed and crossUp, "Source crosses above DSMA", "Source closed above DSMA")
alertcondition(confirmed and crossDown, "Source crosses below DSMA", "Source closed below DSMA")
  • The first two price differences and missing filter states are zero. DSMA follows the source internally during warm-up, rather than starting the average at price zero.
  • The first visible value is at bar index N + 1, the N + 2nd loaded bar. This is a defined startup rule, not proof that the recursive filter has fully settled. Compare later bars and repeat with more history.
  • After warm-up, an exactly zero RMS sets alpha to zero and holds the previous DSMA. Near-zero RMS must be judged together with the numerator; it does not by itself imply a small alpha.
  • The Data Window exposes alpha and RMS. The optional cap is off by default. If you enable it, a raw alpha above 1 becomes 1 and moves DSMA directly to the source.

Both alert conditions require a confirmed chart bar and a valid previous DSMA bar. Create an alert from the desired condition after adding the indicator. TradingView stores an alert snapshot; recreate it after changing the script or its settings. See bar-state rules and alert behavior.

DSMA, KAMA, and VIDYA adapt to different inputs

DSMA scales a filtered price-change signal. KAMA uses an efficiency ratio: net price movement divided by the sum of absolute bar-to-bar moves, then a squared blend of fast and slow smoothing factors. The CMO-based VIDYA documented by MetaTrader weights an EMA factor by the absolute Chande Momentum Oscillator. Equal period labels do not make their response speeds equal.

Compare a line before testing a strategy

  1. Record the exact source, symbol, timeframe, session, adjustment settings, length, filter poles and period, denominator, seed, and cap.
  2. Compare closed-bar DSMA, filter scale, and alpha values after warm-up. Test a flat sequence, a step, a steady ramp, and alternating moves. A line starting near zero is a seed issue to investigate.
  3. Only then specify entries, exits, sizing, costs, and slippage for a strategy test. A crossover indicator alone cannot establish win rate or expected return.

Educational indicator analysis only, not investment advice. The examples are calculations, not evidence of trading performance.

Turn a verified DSMA specification into Pine

Pineify's Pine Script AI Coding Agent can generate and edit code from your description. Specify the RMS formula, seed, filter period, and alpha cap explicitly, then ask it to check syntax. Compare the resulting values in TradingView before adding alerts. Syntax validation does not establish formula equivalence or trading performance.

Build an RMS-based Ehlers DSMA with N = 40, a two-pole filter at N/2, and the startup rules in this example. Show alpha and RMS in the Data Window. Keep the alpha cap optional and off by default. Check syntax, then add closing-bar crossover conditions after numerical comparison.
Explore Pine Script AI Coding Agent

Use public formulas or code you are entitled to modify. A named indicator request does not grant access to proprietary source.

DSMA questions

EDSMA commonly adds Ehlers to the Deviation-Scaled Moving Average name. The acronym alone does not identify an exact implementation. Check the filter period, number of poles, RMS versus standard deviation, seed, and any alpha cap before comparing two lines.

No. Alpha depends on the absolute current filtered value divided by its recent scale. Doubling both values leaves alpha unchanged. If the denominator rises while the current filtered value stays the same, alpha falls. A sudden price move can also take time to pass through the filter.

This example sets alpha to zero when RMS is zero after warm-up, so DSMA holds its previous value. It follows the selected source internally before warm-up and hides that startup segment. Other scripts may retain an old scaled value or use a different fallback; compare their code.

The example uses current and past chart data, with no future references or higher-timeframe requests. The open-bar value can change as the source updates. Its crossover alert conditions require the chart bar to close. A source from another indicator can introduce that indicator's own behavior.

There is no universal best period. The RMS example starts at 40 to match the published demonstration setting, not a trading recommendation. Changing it also changes the Super Smoother period to half that value. Compare response, startup stability, and signals on your chosen symbol and timeframe.

Sources and implementation reviewed October 17, 2026. Continue with the TradingView indicators hub or Zero Lag EMA guide.