How to Backtest a Simple Moving Average (SMA) Strategy

The Simple Moving Average (SMA) remains one of the foundational building blocks of technical analysis and quantitative trading. An SMA calculates the unweighted arithmetic mean of a specified set of price data over a predetermined number of periods. By smoothing out short-term price fluctuations, the indicator provides traders with a clearer view of the underlying market trend.

Mathematical Definition of the SMA

Formally, the Simple Moving Average at time t over a lookback window of n periods is defined as:

Where:

• P_t-i represents the asset price (typically the closing price) at period t-i.
• n represents the chosen lookback period (e.g., 20, 50, or 200 bars).

The Mechanics of SMA Crossover Strategies

While a single SMA can act as a dynamic support or resistance line, single-indicator systems often suffer from high noise sensitivity. To create a robust, actionable trading system, quantitative traders frequently employ a Dual Moving Average Crossover framework.

A dual SMA strategy utilizes two distinct moving averages:

1. Fast Moving Average (SMAfast): Uses a shorter lookback period (e.g., 10 or 20 bars) and responds rapidly to recent price changes.
2. Slow Moving Average (SMAslow): Uses a longer lookback period (e.g., 50 or 200 bars) and reflects the broader macro trend.

The interaction between these two curves generates definitive trading signals:

• Bullish Signal (Golden Cross): Occurs when the SMAfast crosses above the SMAslow. This indicates that recent momentum is accelerating faster than the long-term baseline, signaling a potential long entry.
• Bearish Signal (Death Cross): Occurs when the SMAfast crosses below the SMAslow. This indicates deteriorating short-term momentum relative to the long-term trend, signaling a short entry or position exit.

Backtesting is the practice of applying a mathematical trading model to historical market data to determine how the rule set would have performed in the past. Without rigorous backtesting, live execution is equivalent to speculative gambling.

Key Objectives of Backtesting

1. Quantitative Edge Verification: Proving whether the SMA strategy generates a positive expected value (E > 0) after accounting for market friction.
2. Risk and Drawdown Quantification: Identifying worst-case loss scenarios, maximum peak-to-trough declines, and continuous loss durations.
3. Parameter Sensitivity Analysis: Ensuring the strategy is resilient across different parameters and not hyper-sensitive to exact moving average periods.
4. Behavioral Baseline: Providing the trader with psychological confidence during live execution, reducing panic-driven manual interventions during standard drawdowns.

To obtain realistic and reliable backtest results, quantitative researchers follow a structured execution workflow. Skipping steps or introducing structural shortcuts inevitably leads to inflated expectations and real-world capital losses.

Step 1: Historical Data Acquisition and Cleaning

The validity of any backtest depends entirely on data hygiene. Common challenges in historical market datasets include:

• Missing Bars and Gaps: Unplanned exchange downtime or low liquidity periods can create artificial price jumps.
• Survivorship Bias: Testing only on currently active tokens or equities ignores assets that went bankrupt or were delisted, artificially inflating returns.
• Lookahead Bias: Unintentionally utilizing information in calculation loops that was not available at signal generation time (e.g., using current bar's closing price before the bar closes).
• Timezone Standardization: Mixing UTC timestamps with exchange-local time leads to misaligned candle calculations.

For crypto and forex markets, continuous 1-minute or 1-hour OHLCV (Open, High, Low, Close, Volume) data formats are standard for medium-frequency backtests.

Step 2: Signal Generation Logic

Signal generation requires calculating the moving averages across the dataset and recording state transitions.

Let St represent the binary or scalar trading state at time t:

The actual trade entry or exit trigger occurs on the open of bar t+1, preventing lookahead bias.

Step 3: Modelling Friction (Fees and Slippage)

In a theoretical backtest without friction, SMA strategies on volatile assets often show steady gains. In live trading, transaction costs can erode profitability completely.

1. Exchange Fees: Every trade incurs maker or taker fees (e.g., 0.02% to 0.075% per order).
2. Slippage: The difference between the signal price and the executed price due to order book depth and execution latency.
3. Funding Fees: In crypto perpetual futures, holding leveraged positions incurs dynamic funding rate debits or credits every 8 hours.

A realistic backtest must subtract these costs from every executed trade.

Below is a complete, production-grade Python script utilizing pandas and numpy to run a vectorized backtest for a standard 20/50 SMA crossover strategy on historical daily closing prices.

import numpy as np
import pandas as pd


def run_sma_backtest(
    df: pd.DataFrame,
    fast_period: int = 20,
    slow_period: int = 50,
    initial_capital: float = 10000.0,
    trading_fee: float = 0.0006,  # 0.06% fee per trade
    slippage: float = 0.0002,  # 0.02% estimated slippage
) -> pd.DataFrame:
    """Calculates vectorized backtest performance for a Dual SMA strategy.

    Expects df with a 'close' price column.
    """
    data = df.copy()

    # 1. Compute Moving Averages
    data["sma_fast"] = data["close"].rolling(window=fast_period).mean()
    data["sma_slow"] = data["close"].rolling(window=slow_period).mean()

    # 2. Determine Raw Position Signal (1 for Long, -1 for Short/Cash)
    # Signal generated at close of period t
    data["raw_signal"] = 0
    data["raw_signal"] = np.where(data["sma_fast"] > data["sma_slow"], 1, -1)

    # 3. Shift signal by 1 bar to execute on open/close of t+1 (Eliminating Lookahead Bias)
    data["position"] = data["raw_signal"].shift(1).fillna(0)

    # 4. Calculate Market Returns
    data["market_return"] = data["close"].pct_change().fillna(0)

    # 5. Calculate Raw Strategy Return before friction
    data["strategy_return_raw"] = data["position"] * data["market_return"]

    # 6. Detect Trade Executions (Position Changes)
    data["trades"] = data["position"].diff().fillna(0).abs()

    # 7. Apply Friction Penalties (Fee + Slippage per trade)
    total_friction = trading_fee + slippage
    data["friction_cost"] = data["trades"] * total_friction

    # 8. Net Strategy Return
    data["strategy_return_net"] = data["strategy_return_raw"] - data["friction_cost"]

    # 9. Compute Cumulative Equity Curves
    data["equity_market"] = initial_capital * (1 + data["market_return"]).cumprod()
    data["equity_strategy"] = initial_capital * (1 + data["strategy_return_net"]).cumprod()

    return data


# Example Usage Setup
if __name__ == "__main__":
    # Create mock price series for demonstration
    np.random.seed(42)
    dates = pd.date_range("2023-01-01", periods=500, freq="D")
    price_changes = np.random.normal(0.0005, 0.02, size=500)
    price_path = 100.0 * np.exp(np.cumsum(price_changes))

    df_sample = pd.DataFrame({"close": price_path}, index=dates)
    results = run_sma_backtest(df_sample, fast_period=20, slow_period=50)

    print("Final Strategy Equity: ${:.2f}".format(results["equity_strategy"].iloc[-1]))
    print("Final Market Equity:   ${:.2f}".format(results["equity_market"].iloc[-1]))

Relying purely on total net return is a major mistake in quantitative analysis. A strategy producing a 100% return with a 70% drawdown is far more dangerous than a strategy producing a 45% return with an 8% drawdown.

1. Risk-Adjusted Returns

• Sharpe Ratio: Measures excess return per unit of total risk (standard deviation of returns).
  Sharpe Ratio=

  E[R_p − R_f]σ_p

  An annualized Sharpe Ratio above 1.0 is considered good; above 2.0 is exceptional.

• Sortino Ratio: Similar to Sharpe, but only penalizes downside volatility, ignoring upside spikes.
  Sortino Ratio=

  E[R_p − R_f]σ_d
• Calmar Ratio: Measures the return relative to maximum drawdown.
  Calmar Ratio=

  CAGRMaximum Drawdown

2. Drawdown & Capital Preservation Metrics

• Maximum Drawdown (MDD): The largest percentage drop from a peak to a trough in the equity curve.
  MDD=

  Peak Value − Trough ValuePeak Value
• Drawdown Duration: The time required for the equity curve to recover to a new all-time high following a drawdown phase.

3. Trade Execution Metrics

• Win Rate (Hit Ratio): Percentage of closed trades that resulted in a positive return.
• Profit Factor: Gross profits divided by gross losses.
  Profit Factor=

  ∑ Profits∑ |Losses|

  A Profit Factor above 1.5 indicates a healthy system.

• Expectancy: The average dollar amount or percentage you expect to win or lose per trade.

Even clean code can yield deceptive results if underlying assumptions fail to reflect live trading conditions.

Over-Optimization and Curve Fitting

Traders often test thousands of parameter combinations (e.g., testing fast periods from 1 to 50 and slow periods from 20 to 200) and pick the single pair that yielded the maximum historical return. This practice, known as data snooping, fits the parameters to historical noise rather than persistent market inefficiencies. When deployed live, curve-fitted strategies almost universally fail.

Solution: Use Walk-Forward Analysis or Out-of-Sample (OOS) testing. Split historical data into training sets (70%) and testing sets (30%). Parameter optimization must occur exclusively on the training set, and validated on the out-of-sample set without further adjustment.

Sideways Market Whipsaws

Simple Moving Averages are lagging indicators designed to capture directional trends. During range-bound or sideways market regimes, fast and slow SMAs repeatedly cross each other in rapid succession. This generates multiple false break-out signals, causing capital depletion through repeated "whipsaw" losses and transaction fees.

To improve performance in sideways regimes and reduce false entries, quantitative traders combine SMA crossovers with complementary technical tools.

1. Volatility Expansion Filters (ATR Integration)

Before entering an SMA crossover trade, require the Average True Range (ATR) to be expanding or above a historical percentile. High volatility confirms that a genuine directional breakout is taking place.

2. Trend Confirmation via Directional Indicators (ADX)

Integrating the Average Directional Index (ADX) helps filter out low-trend environments:

• ADX > 25: Strong trend presence. SMA signals are valid.
• ADX < 20: Weak trend or ranging market. SMA signals should be ignored or paused.

3. Multi-Timeframe Alignment

Validate the local signal with higher-timeframe momentum:

• Only take Long SMA crossover signals on the 1-hour chart if the 200-period SMA on the 4-hour or daily chart is sloping upward.
• Only take Short SMA crossover signals if the higher timeframe is in a downward trend.