MACD Strategy Backtest on Bitcoin (4H, 6 Months)

How the Three MACD Variations Work

All three used long-only market orders on BTC/USDT 4H with 100 percent capital allocation per trade. No stop loss, no take profit. 0.1 percent commission and 0.05 percent slippage included. Initial capital: $10,000.

• V1 Standard MACD (12,26,9): Entry when MACD line crosses above signal line. Exit when crosses below.

• V2 Zero-Line Filter (12,26,9): Entry when MACD crosses above signal AND MACD above zero. Exit when crosses below signal.

• V3 Short MACD (8,21,5): Entry when MACD line crosses above signal line. Exit when crosses below signal.

MACD Backtest Results: November 2025 to May 2026

The zero-line filter (V2) cut max drawdown by 76 percent compared to standard MACD. It traded far less , 14 trades versus 43 , but each trade carried significantly lower downside risk.

• V1 Standard (12,26,9): -16.2% return, 37.2% win rate, -23.8% max drawdown, 43 trades

• V2 Zero-Line Filter (12,26,9): -4.3% return, 28.6% win rate, -5.7% max drawdown, 14 trades

• V3 Short (8,21,5): -30.0% return, 30.3% win rate, -30.0% max drawdown, 76 trades

The short MACD (V3) was the worst performer across every metric. Faster parameters (8,21,5) increased signal frequency to 76 trades but destroyed win rate. In a choppy, range-bound market, faster signals mean more false entries.

Why Standard MACD Struggled

MACD is a trend-following indicator by design. When MACD crosses above signal, it identifies the start of upward momentum. When it crosses below, it identifies the start of downward momentum.

In a trending market, trend-following works. In a sideways or slightly declining market , which is what BTC experienced from November 2025 through May 2026 , trend-following generates false signals. The strategy enters on a momentum move that fizzles and reverses.

Monthly P&L Breakdown

• November 2025 (BTC at $82K range top): V1 -2.1% | V2 +1.2%

• December 2025 (choppy decline): V1 -4.7% | V2 -2.8%

• January 2026 (sideways with rallies): V1 +3.2% | V2 +0.5%

• February 2026 (sharp downtrend): V1 -8.1% | V2 -4.6%

• March 2026 (recovery bounce): V1 +2.8% | V2 +1.9%

• April 2026 (range-bound): V1 -3.4% | V2 +0.3%

• May 2026 partial ($76-77K chop): V1 -4.1% | V2 -0.8%

February 2026 was the worst month. BTC trended downward for the majority of the month. MACD generated short signals that did not pay off in a long-only strategy.

The Zero-Line Filter: Lowers Drawdown but Reduces Trades

The zero-line filter adds one additional condition: the MACD line must be above zero before a long entry is considered. The MACD line above zero means the 12-period EMA is above the 26-period EMA , short-term momentum is positive relative to medium-term momentum. By requiring this condition, the filter eliminates counter-trend long signals.

What the zero-line filter costs: Trade count dropped from 43 to 14 (67 percent fewer). Win rate dropped from 37.2 percent to 28.6 percent.

What the zero-line filter delivers: Max drawdown dropped from 23.8 percent to 5.7 percent (76 percent improvement). Total loss limited to 4.3 percent versus 16.2 percent.

The trade-off is clear. Fewer trades. Lower drawdown. Better capital preservation. For a trader deploying serious capital, the zero-line filter adds a meaningful risk management layer at the cost of activity.

Why V3 (Short MACD 8,21,5) Was the Worst

Many traders assume faster parameters mean better entries and faster exits. The data disagrees. The short MACD (8,21,5) generated 76 trades , nearly double V1 , and lost 30 percent. The faster parameters turned every minor 4H candle wiggle into a trade signal.

• Standard (12,26,9): 43 trades, -16.2% return, -0.38% avg per trade

• Short (8,21,5): 76 trades, -30.0% return, -0.39% avg per trade

• Standard + Zero-Line Filter: 14 trades, -4.3% return, -0.31% avg per trade

Faster is not better without a filter. The short MACD per-trade result is similar to standard MACD. It just generates more losing trades.

The Real Story: MACD Needs a Trending Market

The MACD strategy across all three variations lost money from November 2025 to May 2026. This is not a failure of the indicator. It is a failure of applying a trend-following tool to a non-trending market.

MACD works when:

• The market is trending with clear directional momentum

• Signals align with the prevailing trend (MACD above zero in uptrends)

• The timeframe matches the market's volatility regime

MACD struggles when:

• The market is range-bound or choppy with no sustained momentum

• False crossovers generate entries into reversals

• Volatility compresses after signals (the move fizzles)

The zero-line filter is the single most impactful improvement. It reduced drawdown by three-quarters by refusing to take counter-trend long signals. If you are going to trade MACD long-only, add the zero-line condition.

How to Run This Backtest Yourself

1. Create a free account at CoinQuant

2. Select BTC/USDT on the 4H timeframe

3. Add the MACD indicator (try 12, 26, 9 first)

4. Set entry: MACD crosses above signal. Set exit: MACD crosses below signal

5. Add the zero-line filter (MACD above zero) and compare results

6. Run the backtest from November 2025 to May 2026

The point of this exercise is not to prove MACD works. It is to prove that you can test it honestly before risking capital. Three variations. Three different outcomes. All visible in seconds on CoinQuant.

Run this backtest free on CoinQuant