SMARTFINANCEDATA

Statistical Analysis for Traders

Mean Absolute Deviation vs Standard Deviation

Master the key differences between MAD and SD for better statistical analysis and smarter trading decisions.

Volatility Measurement

Understand data dispersion

Outlier Sensitivity

Choose the right metric

Trading Applications

Practical implementation

The Formulas

Mean Absolute Deviation

MAD = Σ|xᵢ - μ| / n

Standard Deviation

SD = √[Σ(xᵢ - μ)² / n]

Start Learning

12 min read

Intermediate Level

8,300+ learners

Understanding MAD vs Standard Deviation in Trading

When analyzing trading performance and market volatility, two fundamental statistical measures emerge: Mean Absolute Deviation (MAD) and Standard Deviation (SD). While both quantify data dispersion, they handle outliers differently and serve distinct purposes in trading analysis. Understanding when to use each metric can dramatically improve your risk assessment and strategy evaluation.

What Are MAD and Standard Deviation?

Both metrics measure how spread out data points are from the mean, but they calculate this dispersion using different mathematical approaches that produce notably different results, especially with volatile or outlier-prone data.

Mean Absolute Deviation (MAD)

MAD = Σ|xᵢ - μ| / n

xᵢ = Each data point

μ = Mean of dataset

n = Number of observations

Standard Deviation (SD)

SD = √[Σ(xᵢ - μ)² / n]

xᵢ = Each data point

μ = Mean of dataset

n = Number of observations

5 Key Differences Between MAD and SD

1. Outlier Sensitivity

The most critical difference lies in how each metric handles extreme values. Standard Deviation squares deviations, which amplifies the impact of outliers dramatically. MAD uses absolute values, treating all deviations proportionally.

Trading Example: Consider daily returns of 1%, 2%, 1.5%, 2%, and 15% (flash spike). The SD will be heavily influenced by the 15% outlier, potentially misrepresenting typical volatility. MAD provides a more stable measure of everyday price movement.

2. Mathematical Properties

Standard Deviation is differentiable and works seamlessly with statistical theory (normal distributions, confidence intervals). MAD is simpler mathematically but less amenable to advanced statistical inference. SD is the foundation for metrics like Sharpe Ratio and Bollinger Bands.

3. Computational Complexity

MAD is computationally simpler, requiring only absolute value operations. SD involves squaring and square root calculations, which can impact performance in high-frequency trading systems processing millions of calculations.

Performance Note: In backtesting systems analyzing thousands of trades, MAD calculations can be 15-20% faster than SD calculations, though modern computers make this difference negligible for most applications.

4. Interpretation and Scale

MAD is in the same units as your data and represents the average deviation. SD is also in the same units but typically larger than MAD due to the squaring effect. For normal distributions, SD ≈ 1.25 × MAD, providing a useful conversion benchmark.

5. Industry Adoption

Standard Deviation dominates finance due to its role in portfolio theory, options pricing (Black-Scholes), and risk management (VaR). MAD is gaining traction in robust statistics and machine learning applications where outlier resistance matters more than mathematical elegance.

Implementing MAD and SD in MQL5

Here's practical code to calculate both metrics and compare their behavior with real trading data:

MQL5 Example: MAD vs Standard Deviation Calculator

// Calculate Mean Absolute Deviation
double CalculateMAD(double data[], int period)
{
   if(period <= 0) return 0.0;
   
   // Calculate mean
   double sum = 0.0;
   for(int i = 0; i < period; i++)
   {
      sum += data[i];
   }
   double mean = sum / period;
   
   // Calculate sum of absolute deviations
   double absDeviationSum = 0.0;
   for(int i = 0; i < period; i++)
   {
      absDeviationSum += MathAbs(data[i] - mean);
   }
   
   return absDeviationSum / period;
}

// Calculate Standard Deviation
double CalculateSD(double data[], int period)
{
   if(period <= 0) return 0.0;
   
   // Calculate mean
   double sum = 0.0;
   for(int i = 0; i < period; i++)
   {
      sum += data[i];
   }
   double mean = sum / period;
   
   // Calculate sum of squared deviations
   double squaredDeviationSum = 0.0;
   for(int i = 0; i < period; i++)
   {
      squaredDeviationSum += MathPow(data[i] - mean, 2);
   }
   
   return MathSqrt(squaredDeviationSum / period);
}

// Compare both metrics with example data
void OnStart()
{
   // Example 1: Normal trading day returns (%)
   double normalReturns[10] = {0.5, -0.3, 0.7, -0.4, 0.6, -0.2, 0.4, -0.5, 0.3, -0.1};
   
   double mad1 = CalculateMAD(normalReturns, 10);
   double sd1 = CalculateSD(normalReturns, 10);
   
   Print("=== Normal Trading Day ===");
   Print("MAD: ", DoubleToString(mad1, 4), "%");
   Print("SD: ", DoubleToString(sd1, 4), "%");
   Print("SD/MAD Ratio: ", DoubleToString(sd1/mad1, 2));
   
   // Example 2: Day with flash crash outlier
   double outlierReturns[10] = {0.5, -0.3, 0.7, -0.4, 0.6, -8.5, 0.4, -0.5, 0.3, -0.1};
   
   double mad2 = CalculateMAD(outlierReturns, 10);
   double sd2 = CalculateSD(outlierReturns, 10);
   
   Print("\n=== Day With Flash Crash ===");
   Print("MAD: ", DoubleToString(mad2, 4), "%");
   Print("SD: ", DoubleToString(sd2, 4), "%");
   Print("SD/MAD Ratio: ", DoubleToString(sd2/mad2, 2));
   
   // Show impact of outlier
   Print("\n=== Outlier Impact Analysis ===");
   Print("MAD increased by: ", DoubleToString((mad2/mad1 - 1)*100, 1), "%");
   Print("SD increased by: ", DoubleToString((sd2/sd1 - 1)*100, 1), "%");
   
   // Recommendation based on data characteristics
   double skewness = (sd2/mad2) / 1.25; // 1.25 is theoretical ratio for normal distribution
   Print("\n=== Recommendation ===");
   if(skewness > 1.2)
      Print("Data has significant outliers - consider using MAD for robust analysis");
   else if(skewness < 0.8)
      Print("Data is more uniform than normal - SD may underestimate risk");
   else
      Print("Data is approximately normal - both metrics are suitable");
}

When to Use Each Metric

Scenario	Use MAD	Use SD
Flash crashes / spikes	✓ Recommended	Overstates risk
Sharpe Ratio calculation	Non-standard	✓ Required
High-frequency data	✓ More stable	Can be noisy
Normal market conditions	Both suitable
Portfolio optimization	Limited theory	✓ Industry standard
Quick volatility check	✓ Faster compute	More precise

Practical Trading Applications

Position Sizing with MAD

Use MAD for position sizing during volatile periods with frequent outliers. It provides a more stable risk estimate that won't cause you to dramatically reduce position sizes after single extreme events.


                            risk_per_trade = account_equity * 0.02;

                            volatility = CalculateMAD(recent_returns, 20);

                            position_size = risk_per_trade / volatility;

Stop Loss Placement with SD

Use Standard Deviation for stop loss placement in trending markets. The sensitivity to outliers helps you set stops that account for worst-case scenarios and avoid premature exits.


                            volatility = CalculateSD(price_changes, 14);

                            stop_distance = 2.0 * volatility; // 2 SD rule

                            stop_loss = entry_price - stop_distance;

EA Performance Monitoring

Calculate both metrics monthly to detect market regime changes. A widening gap between SD and MAD signals increasing outlier frequency, potentially indicating market stress or strategy degradation.

Real-World Example: Comparing Metrics

Let's analyze how MAD and SD behave differently with actual trading scenarios:

Scenario Analysis

Case 1: Steady Trend (Low Volatility)

Daily returns: 0.3%, 0.4%, 0.2%, 0.5%, 0.3%, 0.4%, 0.2%

MAD: 0.11%

Represents typical deviation

SD: 0.13%

Similar to MAD (SD/MAD ≈ 1.18)

Both metrics agree - use either for analysis

Case 2: News Spike Day

Daily returns: 0.3%, 0.4%, -5.2%, 0.5%, 0.3%, 0.4%, 0.2%

MAD: 0.89%

Moderate increase (8x)

SD: 2.07%

Large increase (16x)

SD heavily penalizes outlier - MAD better represents "normal" volatility

Case 3: High Volatility Market

Daily returns: 1.2%, -1.8%, 2.1%, -1.5%, 1.7%, -2.0%, 1.4%

MAD: 1.54%

Higher base volatility

SD: 1.73%

Still proportional (SD/MAD ≈ 1.12)

Both capture volatility well - consistent ratio suggests normal distribution

Advanced Implementation: Hybrid Approach

For robust trading systems, combine both metrics to get comprehensive risk assessment:

MQL5 Example: Hybrid Volatility Monitor

// Hybrid volatility assessment for EA risk management
struct VolatilityMetrics
{
   double mad;
   double sd;
   double ratio;        // SD/MAD ratio
   string regime;       // Market regime classification
   double adjustedVol;  // Risk-adjusted volatility metric
};

VolatilityMetrics AssessVolatility(double returns[], int period)
{
   VolatilityMetrics metrics;
   
   // Calculate both metrics
   metrics.mad = CalculateMAD(returns, period);
   metrics.sd = CalculateSD(returns, period);
   
   // Calculate ratio (normal distribution ≈ 1.25)
   if(metrics.mad > 0)
      metrics.ratio = metrics.sd / metrics.mad;
   else
      metrics.ratio = 0;
   
   // Classify market regime based on ratio
   if(metrics.ratio > 1.5)
   {
      metrics.regime = "High Outlier Activity";
      metrics.adjustedVol = metrics.mad * 1.3; // Use MAD with buffer
   }
   else if(metrics.ratio > 1.25)
   {
      metrics.regime = "Normal Distribution";
      metrics.adjustedVol = metrics.sd; // Use SD (standard approach)
   }
   else if(metrics.ratio > 1.0)
   {
      metrics.regime = "Low Volatility/Uniform";
      metrics.adjustedVol = metrics.sd * 1.1; // Slightly increase SD estimate
   }
   else
   {
      metrics.regime = "Anomalous";
      metrics.adjustedVol = MathMax(metrics.mad, metrics.sd);
   }
   
   return metrics;
}

// Apply hybrid approach to position sizing
double CalculatePositionSize(double accountEquity, double riskPercent)
{
   // Get recent returns
   double returns[50];
   int period = 50;
   
   // Populate returns array (simplified - in real code, fetch from history)
   for(int i = 0; i < period; i++)
   {
      // Your code to get actual return data
      returns[i] = 0.0; // Placeholder
   }
   
   // Assess volatility using hybrid approach
   VolatilityMetrics vol = AssessVolatility(returns, period);
   
   // Calculate risk-adjusted position size
   double riskAmount = accountEquity * (riskPercent / 100.0);
   double positionSize = riskAmount / vol.adjustedVol;
   
   // Log decision
   Print("=== Volatility Assessment ===");
   Print("MAD: ", DoubleToString(vol.mad, 4), "%");
   Print("SD: ", DoubleToString(vol.sd, 4), "%");
   Print("Ratio: ", DoubleToString(vol.ratio, 2));
   Print("Regime: ", vol.regime);
   Print("Adjusted Volatility: ", DoubleToString(vol.adjustedVol, 4), "%");
   Print("Position Size: ", positionSize, " lots");
   
   return positionSize;
}

Common Mistakes to Avoid

❌ Using SD blindly with outlier-prone data

Standard Deviation can overstate risk during flash crashes, leading to overly conservative position sizing and missed opportunities.

❌ Ignoring the SD/MAD ratio

The ratio between these metrics tells you about data distribution. Ignoring it means missing early warning signs of regime changes.

❌ Using MAD for standard financial metrics

Sharpe Ratio, Value at Risk, and other industry-standard metrics require Standard Deviation. Don't substitute MAD unless you're creating custom metrics.

❌ Insufficient sample size

Both metrics need adequate data (minimum 30-50 observations) to be meaningful. Too few data points produce unreliable volatility estimates.

Key Takeaways

MAD is more robust to outliers, making it ideal for volatile markets with flash events
Standard Deviation is required for industry-standard metrics like Sharpe Ratio and VaR
The SD/MAD ratio (normally ~1.25) indicates data distribution and outlier presence
Hybrid approaches combining both metrics provide the most robust risk assessment
Monitor both metrics over time to detect market regime changes and strategy degradation

"The best traders don't rely on a single metric. Understanding when to use MAD versus SD separates sophisticated risk managers from the rest."

Quick Reference Guide

Choose MAD When:

• Data contains frequent outliers
• You need robust volatility estimates
• Flash crashes distort your analysis
• Working with high-frequency data
• Position sizing in volatile markets

Choose SD When:

• Calculating Sharpe Ratio or Sortino Ratio
• Portfolio optimization and theory
• Data follows normal distribution
• Industry-standard reporting required
• Worst-case scenario planning

Remember: SD/MAD Ratio Interpretation

Ratio < 1.0

Anomalous - investigate data

Ratio ≈ 1.25

Normal distribution - ideal

Ratio > 1.5

Heavy outliers - use MAD