Exponential Moving Average (EMA): Formula, Settings, and EMA vs. SMA

An Exponential Moving Average (EMA) smooths prices while giving more weight to recent observations. It is commonly compared with the Simple Moving Average (SMA), which gives equal weight to every close in its window.

EMA is often described as faster. That is useful shorthand for how it responds to a fresh price change, but it does not tell you how often a complete trading rule will act. We examined 97,216 historical configuration evaluations, then added 6,272 fixed-amount activity controls, to separate the weighting calculation from the decision rule built around it.

How EMA is calculated

The usual EMA update is:

alpha = 2 ÷ (length + 1)
EMA now = alpha × close now + (1 − alpha) × EMA previously

For EMA 20, alpha is 2/21, or about 9.52%. Each new close supplies that fraction of the updated average; the previous EMA supplies the rest. SMA 20 instead gives each of its latest 20 closes a 5% weight. TradingView’s moving-average reference describes these calculations.

Initialization matters near the beginning of a series. Different chart packages can start the recursion differently, so two EMA lines may briefly disagree even when they use the same update formula.

The length follows the chart’s candles. An EMA 20 on an hourly chart and an EMA 20 on a daily chart use the same weighting rule over very different elapsed time.

EMA 20 keeps some history older than 20 candles

SMA 20 drops a close completely when it leaves the window. EMA 20 keeps multiplying that close’s influence by 19/21 as new candles arrive. Its influence becomes small rather than disappearing at a fixed age.

With the standard formula, roughly 13.51% of EMA 20’s total weight lies on closes at least 20 candles old. Its latest close receives more weight than SMA 20’s latest close, while its older tail remains present.

There is another useful consequence. Both SMA 20 and the established EMA 20 have a weighted average observation age of 9.5 candles. On a perfectly steady linear price rise, both therefore settle at the same distance behind price. They respond differently when the path bends or becomes noisy. “Faster” describes a response pattern; it is not a fixed number of candles saved on every market move.

Common EMA settings and signals

Shorter settings such as 9 or 20/21 emphasize recent movement. Settings such as 50 and 200 describe broader trends. These numbers are conventions, and none defines a universally best trade.

Common uses include:

• Price relative to EMA: price above the average supplies upward trend context; price below supplies downward context.
• EMA slope: a rising or falling line describes how the smoothed price is changing.
• Two-average crossovers: a shorter EMA moving above or below a longer one changes their relative ordering.
• Pullbacks toward EMA: traders watch how price behaves near the line during a trend.

Fidelity’s EMA guide explains these trend and support/resistance interpretations. The line remains a summary of historical prices. A touch or crossing does not establish that a reversal will follow.

Our controlled experiment focuses on price relative to one average. It does not test every crossover or pullback method described above.

The distance before acting changes the result

We compared otherwise identical long-only rules using SMA or EMA at the same length:

No buffer: buy above the average; exit below it.
With buffer: buy above average × (1 + buffer).
             exit below average × (1 − buffer).

A 2% buffer means 2% above the average for entry and 2% below for exit. This is a fixed percentage band, rather than a standard-deviation band. The gap between entry and exit reduces reactions to small movements around the line.

Each point holds the rule and band fixed and changes only the average length. One execution means one buy or one sell. The displayed activity control uses a small fixed amount to reduce the funding effects that can interrupt orders under near-full allocation.

In the displayed comparison, EMA recorded more executions than SMA at 166 of 196 lengths with no buffer. At a 2% buffer, EMA recorded fewer executions at 163 lengths. This was a change across a broad set of neighboring lengths, rather than a single selected setting.

Across all eight historical comparisons, EMA was busier for a majority of lengths with no buffer. With the 2% buffer, it was quieter for a majority in six of eight. The other two retained a busier EMA majority. The decision band changed the comparison frequently, without making the outcome uniform.

The mechanism is worth keeping in view. An EMA can follow a move more closely, leaving price short of the percentage entry hurdle even when it is far enough from the SMA to qualify. Reacting more strongly to recent prices does not guarantee more completed entry-and-exit cycles. The result depends on the price path, period, and decision distance together.

Fewer transactions do not establish a better investment

A wider band can skip a false crossing, enter a trend later, or miss part of it entirely. Changing the smoother can improve a historical return in one setting and worsen it in another.

The broad study therefore retains return, fees, and platform-reported drawdown alongside activity. It is not a contest decided by the quietest line or the highest single return. The fixed-amount controls isolate activity under a different allocation; their low-exposure returns are not comparable with the main strategy’s returns.

For practical research, write the whole rule before comparing averages: price source, length, entry distance, exit distance, order timing, and allocation. Then compare those same rules across different historical conditions. The label “EMA” supplies only one part of that specification.

Try the tested rule

The defaults use EMA 50 with a 0.5% symmetric buffer. Set smoothing to 0 for the matched SMA rule. The script requests 99% equity sizing and a 500-candle startup; set fees and market dates separately in the backtest settings.

//@version=6
strategy("EMA sensitivity laboratory", overlay=true, initial_capital=10000, default_qty_type=strategy.percent_of_equity, default_qty_value=99, process_orders_on_close=false)
length = input.int(50, "Length", minval=2)
buffer = input.float(0.5, "Symmetric buffer percent", minval=0)
smoothing = input.int(1, "Smoothing: 0 SMA, 1 EMA", minval=0, maxval=1)
s = ta.sma(close, length)
e = ta.ema(close, length)
maValue = smoothing == 0 ? s : e
upper = maValue * (1 + buffer / 100)
lower = maValue * (1 - buffer / 100)
if bar_index >= 500
    if strategy.position_size > 0 and close < lower
        strategy.close("Long")
    else if strategy.position_size == 0 and close > upper
        strategy.entry("Long", strategy.long)
plot(maValue, title="Average", color=color.blue)
plot(upper, title="Entry band", color=color.green)
plot(lower, title="Exit band", color=color.orange)