Skip to main content

Simple vs Exponential

Updated Jun 23, 2026 at 8:47pm

Research Draft Medium 1,331 words

A moving average smooths price into a single trailing line that filters noise and renders trend direction. The two dominant variants differ only in how they weight history: a Simple Moving Average (SMA) weights every period in its window equally, while an Exponential Moving Average (EMA) weights recent prices more heavily and decays older ones geometrically. The core tension is a single tradeoff with no free lunch — the EMA reacts faster (less lag) but is therefore noisier (more whipsaw); the SMA is smoother and steadier but lags more. Neither is universally "better"; the honest, frequently-missed point is that for most strategies the type of average matters far less than the period chosen, the regime, and trading costs.

How it's calculated / formed

SMA — the arithmetic mean of the last N closing prices:

SMA = (P_1 + P_2 + ... + P_N) / N

Each of the N closes carries weight 1/N. Example (5-day): (11+12+13+14+15)/5 = 13. When a new bar arrives, the oldest bar drops out of the window entirely.

EMA — a recursive weighting that never fully discards old data, it just decays it. Two steps:

1. Multiplier (smoothing constant): 

   k = 2 / (N + 1)

For a 10-period EMA, k = 2/11 = 0.1818 (18.18%); for a 20-period EMA, k = 2/21 = 0.0952 (9.52%). (StockCharts/ChartSchool.)

2. Recursive update: 

   EMA_today = (Close_today − EMA_yesterday) × k + EMA_yesterday

Equivalently: EMA_today = Close_today × k + EMA_yesterday × (1 − k). The series must be seeded — the conventional seed is the N-period SMA used as the first "EMA_yesterday," after which the recursion takes over. Because the seed influences early values, EMA series from different platforms can differ slightly until enough bars accumulate to wash the seed out.

Weighted Moving Average (WMA) — a third common variant worth knowing. It weights the N prices linearly: in a 5-period WMA the newest close gets weight 5, the next 4, … the oldest 1, divided by the sum of weights (5+4+3+2+1 = 15). The WMA sits between SMA and EMA in responsiveness — more reactive than SMA, generally a touch less reactive than a same-period EMA, and (unlike the EMA) it fully drops data older than N bars.

How it's used in practice

Type selection is a style choice layered on top of the general moving-average uses (trend filter, dynamic support/resistance, crossover signals — covered in the parent Moving Averages node):

  • EMA — when responsiveness matters. Shorter-term and faster-moving traders favor the EMA (and especially short ones like 9/12/20/21) so the line hugs price and turns quickly. The cost is more false flips in choppy tape.
  • SMA — when stability matters. The 50-day and 200-day SMA are the de-facto institutional/longer-term trend references; their lag is a feature there — it suppresses noise and produces the widely-watched "golden cross / death cross" on the slower, more reliable SMA pair. Mean-reversion and "is this stock above its long trend" filters typically use SMA.
  • Mixing types. A common pattern pairs a fast EMA with a slower SMA (e.g., a short EMA against the 200-day SMA) to get a responsive trigger against a stable trend backdrop.

A useful heuristic: the shorter the lookback and the faster the trade, the more an EMA's responsiveness is worth; the longer the lookback and slower the trade, the more an SMA's smoothness pays. Style-specific entry/stop/target mechanics (e.g., pullback-to-EMA entries) live in the Swing Trading branch — defer to it rather than duplicating here.

Adoption, debate & evidence

Both variants are universally available and universally used — the SMA is the default in most charting packages and underlies the headline 50/200 references; the EMA is the default for many short-term and momentum traders. There is no academic or practitioner consensus that one type is superior; StockCharts states plainly that "one is not necessarily better."

The harder, more important question is whether choosing EMA over SMA materially changes edge. The honest answer from the evidence is mostly no — the type is a second- or third-order knob:

  • Vendor and blog backtests routinely show one type beating the other on a given symbol/timeframe, but the winner flips across markets, periods, and regimes. These results are typically in-sample, slippage-light, and not corrected for data snooping (the more variants you test, the more likely the apparent winner is luck).
  • The foundational positive result for MA rules — Brock, Lakonishok & LeBaron (1992), testing SMA-based rules on the DJIA 1897–1986 — reported significant predictability. But Sullivan, Timmermann & White (1999) showed that once you correct for the universe of rules data-mined over that period (their bootstrap "reality check"), the apparent edge shrinks substantially and largely fails to persist out-of-sample. That critique applies to MA rules in general, well before the SMA-vs-EMA sub-choice.
  • Practical erosion: studies of dual-MA crossover systems find that with realistic slippage/costs a large fraction of parameter combinations are unprofitable. Whether the line is an SMA or an EMA is dwarfed by costs, the chosen period, and the trend/chop regime.
  • Crowding: the 50/200-day SMA levels are watched by enough participants to occasionally be partly self-fulfilling around those exact lines — a reason these specific SMA references retain practical relevance independent of any statistical edge in the averaging method.

Bottom line for the system: treat EMA-vs-SMA as a lag/noise preference, not as an edge source. If a backtest's profitability hinges on the MA type (or flips when the period moves 20→22), that is a robustness red flag, not a discovery.

Strengths & limitations

  • EMA strengths: lower lag, earlier turn-recognition, better for short lookbacks and fast trades. EMA weakness: more whipsaw in sideways/choppy markets; more false crossovers.
  • SMA strengths: smoother, fewer false signals, the recognized long-trend benchmark (50/200). SMA weakness: more lag, slower to flag reversals, and a single outlier bar both entering and later leaving the window can jolt the line twice.
  • Shared limitation: all moving averages are lagging trend tools. In ranging markets every type generates costly false signals; no choice of type fixes that. They also say nothing about why price moves.
  • #1 misuse: over-optimizing the average type (and period) to historical data and mistaking the curve-fit for an edge — the classic data-snooping trap. The second most common misuse is applying any MA crossover system in a non-trending regime.

Worked example

Seed a 10-period EMA with the 10-day SMA, then suppose EMA_yesterday = 100.00 and today's close = 105.00. With k = 2/11 = 0.1818: EMA_today = (105.00 − 100.00) × 0.1818 + 100.00 = 100.91. A 10-day SMA over the same window would move only by (new_close − dropped_close)/10 — so on a sharp up-close the EMA steps up noticeably more than the SMA, visibly illustrating its lower lag (and, in choppy tape, its greater jumpiness).

Sources