DEMA: Double Exponential Moving Average Guide

Key Takeaways

• DEMA double exponential moving average is computed as 2 * EMA(n) minus EMA(EMA(n)), not as the average of two EMAs.
• The construction reduces the inherent lag of a plain EMA by an amount roughly equal to the EMA's own lag.
• The most common mistake is interpreting "double" as "average two EMAs together," which produces a slower line, not a faster one.
• DEMA is widely used as the inner stage of TEMA, T3, and several adaptive systems.

What It Is

DEMA is a recursive smoother that takes a single EMA and corrects it for its own lag. Mulloy noticed that an EMA always trails the data, and that the difference between an EMA and an EMA-of-that-EMA is a good estimate of how far it trails. Subtracting that estimate from twice the original EMA produces a line that sits much closer to current price.

The technique is sometimes called "double smoothing with bias correction." The output is on the price scale and behaves like an EMA, just less laggy and somewhat noisier.

The Intuition

Imagine you are following a friend who is walking five steps ahead. If you always stay five steps behind, you can estimate their position by adding five to your own. EMA-on-EMA estimates how far the first EMA is behind the data. Doubling the first EMA and subtracting that estimate brings the output forward to where the data actually is.

The cost is the same as any low-lag method: the bias correction amplifies noise. A bar that spikes will pull both the EMA and the EMA-of-EMA, but the projection 2*A - B reacts more than either input.

How It Works

The published formula, documented by StockCharts, MetaTrader, and TradeStation, is:

EMA1 = EMA(price, n)
EMA2 = EMA(EMA1, n)

DEMA = 2 * EMA1 - EMA2

Both EMAs use the same period n. Standard values are 20 for daily charts (matching the typical Bollinger Band SMA length) and shorter values like 8 or 13 for intraday systems. The EMAs are computed with the standard smoothing constant k = 2 / (n + 1).

Steady-state lag intuition: a plain EMA(n) lags by roughly n/2 bars. DEMA lags by roughly half that, near n/4 bars, at the cost of a noisier line. Compared to a same-length SMA, DEMA is much faster and closer to current price.

A subtle point: the first few DEMA values are unreliable because both EMAs need warmup. Allow at least 3n bars before reading the line on a fresh series.

Worked Example

Use n = 5 on five closes: 100, 102, 104, 103, 110. Suppose at the bar where price prints 110:

• EMA1 (5-period EMA of price) = 105.0
• EMA2 (5-period EMA of EMA1) = 103.4

Compute DEMA:

DEMA = 2 * 105.0 - 103.4
     = 210.0 - 103.4
     = 106.6

The plain EMA1 reads 105.0; the DEMA reads 106.6. DEMA sits closer to the most recent close of 110 because the projection step nudges it forward.

For comparison, a 5-period SMA of the same prices is 103.8 and a 5-period WMA is 105.2. DEMA's 106.6 is the most responsive of the four on the same data. That responsiveness is the entire reason to use it.

Common Mistakes

1. Interpreting "double" as averaging two EMAs. Some retail tutorials describe DEMA as "two EMAs averaged together." That is wrong. DEMA is 2*EMA1 - EMA2. Averaging two EMAs makes the line slower, not faster.

2. Using DEMA on illiquid or gappy series. The bias correction amplifies single-bar spikes. On thin tickers or pre-market data, a gap can pull DEMA much further than it pulls a plain EMA.

3. Choosing too short a period. DEMA reacts roughly like an EMA of half its length. A 5-period DEMA can behave like a 2- or 3-period EMA, which is faster than most strategies actually want.

4. Comparing DEMA and EMA at equal length. They are not equivalent at equal length. Compare a length-n DEMA to a length-2n EMA for similar smoothing, and check signal frequency rather than line shape.

5. Ignoring warmup. Both EMA1 and EMA2 need stabilization. Until they do, DEMA outputs are arbitrary. Skip at least 3n bars at the start of any backtest.

Frequently Asked Questions

What is DEMA double exponential moving average in simple terms? The DEMA double exponential moving average is a smoother that subtracts the lag of an EMA from twice itself, so the line sits closer to current price.

How does DEMA affect investment decisions? Short-term traders use DEMA in place of EMA when they need faster trend confirmation without switching to a much shorter EMA. Portfolio-level applications are rare because DEMA's noise hurts on long horizons.

What is a real-world example of DEMA use? DEMA is a built-in study in StockCharts, TradingView, MetaTrader 5, and TradeStation. It is also the inner building block of TEMA and Tim Tillson's T3 indicator.

How can investors use DEMA effectively? Match the length to your horizon, expect more noise than a same-length EMA, and confirm signals with a longer trend filter or a volatility regime. Use it as one input, not a standalone system.

How is DEMA different from TEMA? DEMA combines a single EMA and an EMA-of-EMA. TEMA adds a third stage, EMA-of-EMA-of-EMA, in the combination 3EMA1 - 3EMA2 + EMA3, which projects the line even closer to price at the cost of more noise.

Sources

1. StockCharts ChartSchool. "Double Exponential Moving Average (DEMA)." https://chartschool.stockcharts.com/table-of-contents/technical-indicators-and-overlays/technical-overlays/double-exponential-moving-average-dema
2. MetaTrader 5 Help. "Double Exponential Moving Average." https://www.metatrader5.com/en/terminal/help/indicators/trend_indicators/dema
3. TradingView Help. "Double Exponential Moving Average (EMA)." https://www.tradingview.com/support/solutions/43000589132-double-exponential-moving-average-ema/
4. TradeStation EasyLanguage. "DEMA Function." https://help.tradestation.com/10_00/eng/tsdevhelp/elword/function/dema_function_.htm

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.