Trend & directionDouble Exponential Moving Average · DEMA
Patrick Mulloy's construction that halves EMA lag by correcting for the smoothing delay.
Works best in trending marketsEngine-computed on a fixed sample series
What it is
The Double Exponential Moving Average, or DEMA, is a moving average engineered to react faster to price by cancelling out much of the lag that a normal average carries. Patrick Mulloy introduced it in 1994 to solve a persistent problem: every moving average trails behind price because it is an average of the past, and that delay costs traders entries and exits. Despite its name, DEMA is not simply an exponential average of an exponential average; it is a specific construction that subtracts the extra smoothing lag back out. The result is a line that hugs current price more closely and turns sooner than an ordinary exponential average of the same length. For a beginner, DEMA answers the question, can I have the smoothness of a moving average without so much of the delay? It is used exactly like any moving average, just with quicker response.
How it is calculated
The formula is twice the exponential moving average of price, minus the exponential moving average of that exponential moving average. In words, you first compute an ordinary EMA of the chosen length, then compute an EMA of that EMA to measure how much lag the smoothing introduced, and then you add that lag estimate back to the first EMA. Because the single EMA lags price and the double-smoothed EMA lags even further, the gap between them approximates the delay, and subtracting the double-smoothed component from twice the single EMA removes most of it. The key point is that DEMA is not an EMA of an EMA, which would be slower, but a lag-corrected combination that is faster than a single EMA. It remains a genuine, well-behaved average rather than a raw price line, just one that tracks more closely.
Reading it, step by step
Read DEMA exactly as you would any moving average, but expect everything to happen sooner. Its slope gives the trend direction, and because it sits nearer to current price than an equal-length EMA, it changes direction earlier at turning points. Crossovers of price through the DEMA, or of a fast DEMA through a slow one, are your signals, and they arrive ahead of the equivalent EMA signals. The line's closeness to price is the visual tell of its reduced lag — it clings to the action rather than trailing well behind. The trade-off you are reading is sensitivity: DEMA responds faster, which means it also reacts more to short-term noise, so a given wobble in price shows up more in DEMA than in a plain EMA. In short, same interpretation, earlier signals, slightly more sensitivity.
Best timeframes and settings
DEMA takes a single length parameter, just like an EMA, and it can be dropped into any system that uses a moving average across any timeframe. Shorter lengths suit active intraday and swing trading where the reduced lag translates into meaningfully earlier entries, while longer lengths still beat a plain EMA of the same length to the turn and suit position traders who want a responsive trend filter. The fundamental trade-off is that less lag means more responsiveness to noise, so a DEMA will chop more in sideways markets than the EMA it replaces. A common approach is to swap DEMA in wherever an EMA is currently costing you entries through delay, keeping the same length. Because it is inherently faster, there is rarely a need to shorten it aggressively; if anything, a slightly longer DEMA can give EMA-like smoothness with less lag.
When and where to use it
DEMA is at its best in trending markets, where its reduced lag lets you enter and exit closer to the actual turns and capture more of each move. It applies to any asset class that uses moving averages — equities, futures, FX, and crypto. In choppy, sideways markets its greater responsiveness becomes a liability, because it reacts to the very noise a slower average would smooth away, producing more false crossovers. Use it to replace an EMA in crossover systems, trend filters, and trailing references where lag is the specific problem you want to reduce. Avoid reaching for it as a standalone signal in ranging conditions, where its sensitivity works against you. The honest framing is that DEMA reduces lag but cannot remove it, and it buys that reduction with extra noise sensitivity that only pays off when there is a real trend to track.
Strategies that use it
The first strategy is a lag-reduced crossover system: replace the EMAs in a fast-slow moving-average crossover with DEMAs of the same lengths, taking long signals when the fast DEMA crosses above the slow and shorts on the reverse, gaining earlier entries than the EMA version would give. The second is a responsive trend filter: use a single longer DEMA as the line that price must be above to permit longs and below to permit shorts, so your directional bias updates sooner when the trend actually changes. The third is a trailing reference: trail a DEMA beneath a long position as a dynamic stop that tracks price more tightly than an EMA, locking in more of a trend while still smoothing minor pullbacks. In each case, confirm signals with a slower average or a momentum tool to offset DEMA's greater noise sensitivity in flat stretches.
Combining it with other indicators
DEMA belongs to the low-lag moving-average family and pairs well with tools that confirm the trend its speed helps you catch early. A slower moving average, whether a plain EMA or a DEMA of longer length, filters DEMA's noisier signals and defines the larger trend the faster line should trade within. The ADX confirms that a trend is genuinely present, which is the condition under which DEMA's reduced lag is an asset rather than a whipsaw generator. A momentum oscillator can corroborate that an early DEMA crossover has real thrust behind it. Related constructions such as the Triple Exponential Moving Average and the T3 average sit in the same low-lag toolkit and can be compared or paired for a fast-slow spread. The recurring idea is to let a slower or confirming tool separate DEMA's genuine early signals from the extra noise its speed introduces.
Where it fails
DEMA's central weakness is the flip side of its strength: less lag means more responsiveness to noise, so it whipsaws more than a plain EMA in sideways markets and can throw off false crossovers where a slower average would stay flat. It reduces lag but cannot eliminate it, so at very sharp reversals it still turns after the actual top or bottom, just less late than an EMA. Beginners sometimes assume DEMA is strictly better than an EMA and use it everywhere, only to get chopped up in ranges where the EMA's extra smoothing would have kept them out. The fixes are to reserve DEMA for trending conditions, to confirm its signals with a slower or momentum-based filter, and to avoid shortening it so aggressively that it becomes little more than a jittery price proxy. Match the tool to the regime: DEMA rewards a clean trend and punishes chop more than the average it replaces.
A worked example
Suppose a stock is trading at 100 in a steady uptrend. A standard 20-period EMA sits at 97, lagging three points behind price, and an EMA of that EMA sits at 95, lagging further. DEMA is twice the first EMA minus the second, or two times 97 minus 95, which is 194 minus 95, equals 99. Notice that DEMA at 99 is much closer to the current price of 100 than the EMA at 97 — that one-point gap versus three points is the lag correction at work, computed as the EMA of 97 plus the two-point difference between the EMA and its double-smoothed version. In a crossover system, this closeness means that when the stock finally rolls over, price will cross below the DEMA a bar or two before it crosses below the slower EMA, giving an earlier exit. The cost is that during a brief sideways wobble the same DEMA will react to the noise and may flip while the EMA holds flat, which is exactly why you would confirm the signal with a longer average.