MACD Strategy — Signals, Settings and Historical Test Results

MACD has a reputation as a "beginner's" indicator: two lines, a crossover up — buy, down — sell. That simplicity is both its strength and its trap, because hardly anyone checks whether those crossovers actually made money historically. And for MACD, the data exists — and it's more interesting than the textbook legend: on the S&P 500 index, it was tested over more than six decades of trading, and the results say plenty about both the strength and the limits of this tool.

What MACD Is

MACD (Moving Average Convergence Divergence) is built from three elements, all calculated from the standard 12/26/9 settings:

• The MACD line = EMA(12) - EMA(26). When it's above zero, the fast average is above the slow one (the market is in a short-term uptrend); a cross through zero is the exact equivalent of those two averages crossing on the price chart.
• The signal line = a 9-period EMA of the MACD line — a smoothed version the MACD "oscillates" against.
• The histogram = the difference between the MACD line and the signal line — it shows whether momentum is building or fading before a crossover happens.

Three classic types of signals:

1. Signal-line crossover — MACD crosses above the signal line (buy signal) or below it (sell signal). The most common and fastest signal, but also the most prone to noise.
2. Zero-line crossover — slower, rarer; confirms a change in the short-term regime. It's a cousin of the golden cross, just on shorter averages.
3. Divergence — price makes a new high while MACD/the histogram doesn't: momentum is fading. A warning signal, not a trade trigger.

MACD is a lagging indicator — built from averages, it's inherently late. It doesn't catch lows and highs; in exchange, it can keep you on the right side of a longer move. That defines its natural environment: trends. And its hell: consolidation.

What the Numbers Say

The longest publicly described MACD test comes from QuantifiedStrategies: the S&P 500 since 1960 — over 60 years of data (numbers from summaries — verify at the source before citing):

Three things are worth reading from this table — and none of them are what the courses suggest.

First: 112 trades over 66 years is fewer than 2 signals a year. A properly tested MACD on the daily timeframe isn't a click-every-day machine — it's a rare regime filter. Anyone generating several MACD signals a week is using it on lower timeframes, where noise dominates, and is playing an entirely different (untested) strategy.

Second: +1.74% average per trade with a 71.5% win rate is a solid, but not spectacular, statistic. A positive expectancy on a large sample over a long period — that's a lot more than most popular indicators can show (the classic RSI 30/70 lost on 2,397 trades). But it isn't a "life-changing edge" — it's a modest, repeatable trend edge.

Third: the BTC version looks spectacular — and that's exactly why it needs caution. A momentum-strategy backtest based on MACD crossovers on Bitcoin reported a CAGR around 77%. Before that number lands in your spreadsheet: it comes from a summary (verify at the source), it covers a period when BTC itself grew exponentially (part of the result is exposure to the bull run, not credit to the indicator), and it needs comparing against buy-and-hold plus checking over a full cycle including the 2022 bear market. Momentum on crypto has independent support in the literature — but treat this specific number as a hypothesis to verify, not a promise.

As always: past results don't guarantee future ones, and every QS number should be confirmed directly on the source page before you make any decision based on it.

How to Apply It Step by Step

A baseline version — MACD as a trend system on the daily, modeled on the logic from the tests:

1. Timeframe and market: daily candles, a liquid market (BTC, ETH, indexes). On M5-M15, MACD generates mostly noise — the tested statistics apply to the daily.
2. Trend filter: only trade long when price is above the SMA200. MACD by itself doesn't distinguish a pullback in a bull market from the start of a bear market — the filter does that job for it and cuts out most of the signals born in consolidation.
3. Entry signal: the MACD line crosses the signal line upward below zero (a bounce after a pullback in an uptrend) or crosses the zero line upward (a slower variant, fewer signals, less noise). Enter on the close of the signal candle.
4. Exit: a downward MACD/signal crossover (the symmetric variant), or a trailing stop below successive swing lows — trailing lets you avoid giving back the whole move on a lagging exit signal.
5. Stop loss: below the last local low with an ATR buffer — a moving-average-based signal tends to lag, so a technical stop protects against a scenario where you never get an "exit" crossover with a meaningful balance left.
6. Risk: 1% of the account per trade; with fewer than 2 signals a year per market, it's worth running a portfolio of several independent markets so the strategy has any sample size to work with.

A numerical example on BTC (illustrative): a $10,000 account, 1% risk = $100. BTC is above the SMA200 (price $98,000, the average at $88,000). After a three-week pullback, the MACD histogram has been rising for four sessions, and on the daily close, the MACD line crosses the signal line upward below zero. Entry at $98,000, stop below the pullback low at $93,500 (risk of $4,500 per BTC) → position size = 100 / 4,500 ≈ 0.022 BTC (a notional of ~$2,180). Management: a trailing stop below successive higher daily lows. If the trend develops to $112,000, the position earns roughly +$311 (3.1R); if the signal turns out false, the loss is capped at $100. Notice the proportion: the result is decided by managing the position, not by the crossover itself.

MACD as a Filter, Not an Oracle

A second, often more sensible application: MACD doesn't generate entries, it confirms the regime for another strategy. Examples: you play breakouts from consolidation only in the direction matching the sign of MACD on a higher timeframe; or you take mean-reversion signals only when the histogram isn't showing strong opposing momentum. Lagging indicators work best exactly this way, as filters — a similar conclusion came out of ADX tests, where it performed mediocrely as a standalone signal but noticeably improved other strategies as a filter.

When It Doesn't Work and Common Pitfalls

• Consolidation is a loss machine. In a sideways market, the MACD lines cross back and forth around zero, generating a series of false signals in both directions. This is a structural flaw of every moving-average system — a trend filter and a higher-timeframe direction requirement are mandatory, not optional.
• Lag on reversals. MACD confirms a move that's already underway. On sharp reversals (crypto crashes), the exit signal from a crossover arrives after a double-digit percentage decline — which is why a hard stop loss exists independently of the indicator.
• Low timeframes. The tested statistics apply to the daily. On M5, MACD is mostly measuring microstructure and noise — its "signals" there have no documented edge, and transaction costs rise.
• Divergences as standalone signals. A MACD divergence can "get it wrong" repeatedly before a trend actually turns — in a strong bull run, momentum can fade for months while price keeps rising. Treat it as a warning to tighten your stop, not an invitation to fade the trend.
• Optimizing parameters to fit a chart. 12/26/9 doesn't work on your pair? Tuning it to 8/21/5 until history looks good is overfitting in its purest form. If the standard settings show no edge in a rigorous test on a given market, the conclusion is "MACD doesn't work here," not "let me look for better parameters."
• Confusing bull-market exposure with a strategy's edge. Any long-only trend system tested on a rising market looks good. An honest test compares the result against buy-and-hold and checks behavior in a bear market — exactly the caveat that applies to the spectacular 77% CAGR on BTC.

And one final, most mundane pitfall: MACD on a chart is not a strategy. An indicator with no written rules for entry, exit, stop and position size is decoration — and decorations have a way of showing everyone exactly what they want to see, especially after the fact.

The "no hype" conclusion: MACD has something most popular indicators don't — a long, measurable history of positive expectancy on indexes. But that statistic applies to rare daily signals with a trend filter, not to daily clicking on the M15. Use it the way it was tested — as a slow trend tool or a regime filter — and check the numbers that convinced you at the source, on your own market, before you put money behind them.

FAQ

Rafał — Krypto Bez Ściemy

Trader and founder of Strefa Tradingu. He’s been breaking crypto down on YouTube for years — no hype, no signals, with a focus on market structure and risk management.

kryptobezsciemy.plInstagramYouTubeKlub Zero