Skip to main content

Moving Averages for Swings

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,180 words

A moving average (MA) is a continuously recomputed average of recent closing prices that smooths noise into a single trend line. For a swing trader holding days to a few weeks, the MA is rarely a standalone signal generator — it is a context and structure tool: it defines whether a name is in an up-, down-, or no-trend, marks dynamic support/resistance that pullbacks tend to respect, and supplies objective entry, stop, and exit anchors. The core tension is lag versus noise: a short MA reacts fast but whipsaws in chop; a long MA filters noise but signals late. Swing trading lives in the middle of that trade-off.

How it's calculated / formed

  • Simple Moving Average (SMA): the arithmetic mean of the last N closes. A 20-day SMA sums the last 20 closes and divides by 20, dropping the oldest value each new day (StockCharts ChartSchool).
  • Exponential Moving Average (EMA): weights recent prices more heavily, reducing lag. Per ChartSchool, the multiplier is 2 / (N + 1) (≈0.0952 for a 20-EMA), and each day: EMA = (Close − prior EMA) × multiplier + prior EMA, seeded with an initial SMA.
  • EMA vs SMA, practically: EMAs turn faster and hug price more closely, giving earlier — but noisier — signals; SMAs are smoother and lag more. ChartSchool notes neither is inherently superior. Swing traders favoring quicker, shorter holds lean EMA (e.g., 9/21); those wanting fewer false signals lean SMA.

Lengths swing traders actually use (a convention, not a law — common across StockCharts, Zerodha Varsity, and broker education):

  • 9- or 10-period (EMA) — fast trigger line for short holds and the edge of a tight pullback.
  • 20-period / 21-EMA — the workhorse "swing line"; healthy uptrends often pull back to it and resume.
  • 50-period (SMA) — the intermediate trend filter; the line institutions and chartists watch most.
  • 200-period (SMA) — the long-term regime line dividing secular bull from bear.

How it's used in practice

Decision-useful patterns a swing trader keys on:

1. Trend filter (the gate). Only take long swings when price and the stack are aligned: price above a rising 20- and 50-MA, ideally with 50 > 200. This is John Murphy's "trade with the trend" principle (StockCharts). The 50- and 200-day act as the on/off switch; the 20 times the entry.

2. Pullback-to-MA entry (the bread-and-butter swing). In an established uptrend, price retraces into the 20-EMA or 50-SMA, holds (a tight reversal candle, intraday rejection wick, or higher low), then resumes. Trigger = reclaim of the prior day's high or a close back above the fast MA. Stop = below the swing low / below the MA being tested. This converts the MA into a zone of value rather than a crossover.

3. Crossover signals (slower, trend-change oriented). A fast MA crossing above a slow MA (e.g., 9/20 EMA, or 20-EMA over 50-SMA) flags a momentum shift. Practitioners (5paisa, Dhan, VectorVest) stress crossovers are far more reliable filtered: a 9/20 cross is meaningful only with a rising 50-EMA beneath it and expanding volume — a raw crossover alone whipsaws.

4. The 200-day as the line in the sand. Many swing traders simply refuse longs below a falling 200-day SMA and refuse shorts above a rising one — a regime gate, not a timer.

5. MA as trailing stop / exit. Riding a winner, exit on a decisive close below the 20-EMA (aggressive) or 50-SMA (giving the swing more room). This systematizes the exit and removes discretion.

Adoption, debate & evidence

Moving averages are arguably the single most widely used technical overlay; the 50- and 200-day are referenced even in mainstream financial media. But adoption ≠ proven edge, and the evidence is genuinely contested:

  • The bull case (in-sample): Brock, Lakonishok & LeBaron (1992, Journal of Finance) tested MA rules on the Dow 1897–1986 and found buy signals produced higher, less-volatile returns than sell signals — results inconsistent with a random walk, AR(1), or GARCH (Wiley). This is the most-cited academic support.
  • The critical case (out-of-sample / data-snooping): Sullivan, Timmermann & White (1999, Journal of Finance) re-tested the same rule universe with a bootstrap "Reality Check" correcting for data-snooping bias, and found the rules' profitability largely vanished in the post-1986 out-of-sample period, interpreting it as evidence of increasing market efficiency (Wiley). The honest read: MA rules looked great precisely because they were the rules people searched and found.
  • Golden cross / death cross folklore vs measured: A widely cited multi-decade S&P 500 backtest reports a high win rate (commonly quoted around ~79%) for the 50/200 golden cross — but the same analyses stress this is on a diversified index that absorbs single-stock failures, produces only a few dozen signals over decades, lags buy-and-hold on raw CAGR, and earns its keep mainly through reduced drawdown / better risk-adjusted return by sitting out bear markets (QuantifiedStrategies, TOSIndicators backtests — vendor sources, treat win-rate figures as indicative not gospel). On individual stocks the same cross fires more often, with more false signals.

Folklore vs measured bottom line: as a trend/regime filter and risk-management overlay, MAs have defensible value. As a standalone crossover signal-generation system, the academic out-of-sample evidence is weak.

Strengths & limitations

  • Works when: the market is trending. ChartSchool: MAs "work brilliantly in strong trends" — the pullback-to-MA swing is most reliable inside a clean, already-established trend.
  • Fails when: the market is ranging/sideways. The same source warns MA systems "produce relatively late signals" and "many whipsaws" in non-trending markets, and that you should never expect to sell tops or buy bottoms with them. Crossovers in chop are a death-by-a-thousand-cuts failure mode.
  • Lag is structural: longer MAs require a larger, longer move to turn — the signal arrives after a chunk of the move is gone.
  • #1 misuse: trading the raw crossover mechanically with no trend filter, no volume confirmation, and no regard for whether the instrument is trending — and treating the round-number 50/200 as magic rather than as a self-fulfilling, widely-watched level.

Sources