Quantitative Risk Management: Position Sizing and Drawdown Control

Quantitative risk management is the discipline that separates surviving traders from failed ones. While entry and exit signals receive the most attention, position sizing and risk management determine whether a strategy produces consistent returns or catastrophic losses. As Ed Seykota famously stated, "There are old traders and there are bold traders, but there are very few old, bold traders." Research by Balsara (1992) demonstrated that optimal position sizing can transform a modestly profitable strategy into a significantly profitable one, while poor sizing can make any strategy unprofitable.

This guide covers the quantitative framework for managing risk at both the individual trade and portfolio level.

Position Sizing Models

Fixed Fractional Method

The most common approach: risk a fixed percentage of equity on each trade.

• Risk amount = $100,000 * 0.01 = $1,000
• Risk per share = $50 - $48 = $2
• Position size = $1,000 / $2 = 500 shares

Volatility-Adjusted Position Sizing

Size positions inversely proportional to their volatility, ensuring equal risk contribution:

This method ensures that a low-volatility stock (e.g., a utility) and a high-volatility stock (e.g., a biotech) contribute approximately equal risk to the portfolio.
| Risk Method | Description | Typical Risk% |
|-------------|-------------|---------------|
| Conservative | 0.5% per trade | Beginners, drawdown-sensitive |
| Moderate | 1.0% per trade | Standard systematic trading |
| Aggressive | 2.0% per trade | High-conviction, experienced |
| Kelly Criterion | Optimal | Maximum growth (see below) |

Kelly Criterion

The Kelly criterion calculates the mathematically optimal bet size to maximize long-term geometric growth:

Where W = win rate and R = average win / average loss ratio.

• Kelly% = (0.55 * 1.5 - 0.45) / 1.5 = 0.25 = 25%

Anti-Martingale Position Sizing

Increase position sizes during winning streaks and decrease during losing streaks. This exploits the fact that winning and losing trades tend to cluster:

• Base position: 1% risk per trade
• After 3 consecutive wins: Increase to 1.5% per trade
• After 3 consecutive losses: Decrease to 0.5% per trade
• Reset: After any position size change, require 3 trades at the new size before further adjustment

Drawdown Analysis and Control

Understanding Drawdowns

A drawdown is the peak-to-trough decline in account equity. It measures the worst loss from any peak before the account reaches a new high.

• Maximum drawdown: The largest peak-to-trough decline in the backtest period
• Average drawdown: The mean of all drawdown periods
• Drawdown duration: How long the account was below its peak
• Recovery time: How long it took to reach a new equity high

Historical Drawdown Expectations

Based on Monte Carlo analysis of a strategy with Sharpe ratio 1.0 and 10% annual volatility:
| Probability | Expected Max Drawdown |
|------------|----------------------|
| 50% chance | -14.2% |
| 25% chance | -18.4% |
| 10% chance | -22.8% |
| 5% chance | -26.1% |
| 1% chance | -32.4% |

Even a good strategy (Sharpe 1.0) has a 5% chance of a drawdown exceeding -26%. Traders must be psychologically and financially prepared for these events.

Dynamic Drawdown Control

Implement automatic risk reduction during drawdowns:
| Current Drawdown | Action |
|-----------------|--------|
| 0% to -5% | Full position sizes |
| -5% to -10% | Reduce to 75% position sizes |
| -10% to -15% | Reduce to 50% position sizes |
| -15% to -20% | Reduce to 25% position sizes |
| > -20% | Stop trading, review strategy |

This approach limits the depth of drawdowns at the cost of slower recovery. In our backtest, dynamic drawdown control reduced maximum drawdown from -22.4% to -15.8% while reducing CAGR by only 1.2%.

Value at Risk (VaR)

Historical VaR

Calculate VaR from the historical distribution of returns:

For a $1 million portfolio with daily VaR of -1.8%, the interpretation is: "There is a 95% probability that the portfolio will not lose more than $18,000 in a single day."

Parametric VaR

Assume returns are normally distributed:

For 95% confidence: Z = 1.645
For 99% confidence: Z = 2.326

Conditional VaR (CVaR / Expected Shortfall)

CVaR measures the expected loss given that VaR is exceeded:

CVaR is a more conservative and coherent risk measure than VaR because it accounts for the severity of tail losses, not just their probability.

VaR Comparison for a Typical Quant Portfolio

CVaR consistently produces the most conservative estimates, making it the preferred measure for risk-conscious portfolios.

Portfolio-Level Risk Management

Correlation-Aware Position Sizing

Positions that are highly correlated should be treated as a single concentrated bet:

1. Calculate the correlation matrix of all portfolio holdings
2. Group holdings with correlation > 0.7 as a single risk unit
3. Apply position size limits to the group, not individual positions

Sector and Factor Limits

Stress Testing

Subject the portfolio to historical and hypothetical stress scenarios:
| Scenario | Portfolio Impact | Action |
|----------|-----------------|--------|
| 2008 Financial Crisis | Replay 2008 returns | Acceptable if < -15% |
| Flash Crash (2010) | -8% in minutes | Test kill switch |
| COVID Crash (March 2020) | -34% in 23 days | Test drawdown controls |
| Interest Rate Shock | +200 bps in 30 days | Check duration exposure |
| Correlation Breakdown | All correlations go to 1.0 | Test diversification failure |

If any scenario produces losses exceeding tolerance, adjust portfolio construction before deploying.

Risk-Adjusted Performance Metrics

Beyond Sharpe Ratio

Use multiple metrics to get a complete picture. A strategy with a high Sharpe but low Calmar has hidden tail risk.

Key Takeaways

• Position sizing has more impact on portfolio performance than entry signals
• Fixed fractional (1% risk per trade) is the recommended starting point for systematic traders
• Half Kelly or quarter Kelly provides a practical balance between growth and drawdown
• Dynamic drawdown control (reducing sizes during drawdowns) reduced max drawdown from -22.4% to -15.8% with only 1.2% CAGR cost
• CVaR (Expected Shortfall) is a more robust risk measure than VaR for fat-tailed return distributions
• Correlation-aware position sizing prevents hidden concentration risk
• Stress testing against historical crises is essential before deploying capital

Frequently Asked Questions

What is the optimal risk per trade for a systematic strategy?

Research and practitioner consensus suggest 0.5-2.0% risk per trade, with 1.0% being the most common choice. This balances growth potential with drawdown control. At 1% risk per trade, even a streak of 10 consecutive losses only produces a 9.6% drawdown (compounding effect). Higher risk percentages accelerate both gains and losses, making them suitable only for high-conviction strategies with proven edge.

How do you calculate maximum drawdown?

Maximum drawdown is calculated by tracking the running maximum of the equity curve and measuring the percentage decline from each peak. For each point in time: Drawdown(t) = (Equity(t) - Max_Equity_to_t) / Max_Equity_to_t. The maximum drawdown is the minimum value of this time series. Always report both the depth and duration of the maximum drawdown, as a -15% drawdown lasting 2 months is very different from a -15% drawdown lasting 18 months.

Is the Kelly criterion practical for trading?

Full Kelly is impractical because it produces extreme drawdowns (50%+ is common) and requires perfectly accurate estimates of win rate and payoff ratio, which are unknown in practice. However, fractional Kelly (typically half or quarter Kelly) provides a useful framework for position sizing by linking bet size to edge magnitude. Use Kelly as an upper bound, then reduce based on parameter uncertainty and risk tolerance.

How do you manage risk across multiple strategies?

Portfolio-level risk management for multiple strategies requires: (1) calculate correlation between strategies (low correlation = better diversification), (2) allocate capital inversely proportional to strategy volatility (volatility parity), (3) set portfolio-level limits (max drawdown, max gross exposure) in addition to strategy-level limits, (4) monitor aggregate factor exposure to prevent hidden directional bets. A portfolio of 3-5 uncorrelated strategies with Sharpe ratios of 1.0 each can produce a combined Sharpe of 2.0+ through diversification.