Understanding Maximum Drawdown
Maximum Drawdown (MDD) quantifies the largest peak-to-trough decline in an investment's value. It measures downside risk. MDD represents the worst historical loss an investor would have endured. Calculating MDD involves identifying the highest point (peak) before a new low (trough). The percentage drop from that peak to that trough is the MDD.
Consider a mean reversion strategy. It starts with $1,000,000. Month 1: Portfolio value increases to $1,050,000. Month 2: Portfolio value drops to $1,020,000. Month 3: Portfolio value rises to $1,070,000. Month 4: Portfolio value falls to $980,000. Month 5: Portfolio value recovers to $1,000,000.
The highest peak before a new low was $1,070,000 in Month 3. The subsequent trough was $980,000 in Month 4. MDD = (($1,070,000 - $980,000) / $1,070,000) * 100% = 8.41%.*
MDD is a backward-looking metric. It reflects past performance. It does not predict future drawdowns. However, it provides a realistic expectation of potential losses. Traders use MDD for risk budgeting. It informs position sizing and capital allocation.
Mean Reversion and Drawdown Dynamics
Mean reversion strategies often exhibit distinct drawdown characteristics. These strategies profit from price deviations from a historical average. They typically involve frequent, smaller trades. This often leads to a higher win rate but smaller profit per trade. When prices trend strongly against the mean, these strategies can experience significant drawdowns.
Consider a simple pairs trading strategy. It involves two highly correlated stocks, say MSFT and AAPL. The strategy buys the underperforming stock and sells the outperforming stock when their price spread deviates significantly. It closes positions when the spread reverts to its mean.
Imagine a pairs trading strategy on MSFT and AAPL from January 1, 2020, to December 31, 2020. Initial Capital: $1,000,000. January 2020: Strategy profits, portfolio grows to $1,020,000. February 2020: Market volatility increases. The MSFT-AAPL spread widens significantly. The strategy experiences losses. Portfolio drops to $950,000. This is a drawdown of (($1,020,000 - $950,000) / $1,020,000) * 100% = 6.86%. March 2020: Market continues to trend. The spread diverges further. Portfolio drops to $880,000. The new peak for MDD calculation is $1,020,000. The trough is $880,000. MDD = (($1,020,000 - $880,000) / $1,020,000) * 100% = 13.73%. April 2020: The spread begins to revert. Strategy profits. Portfolio recovers to $920,000. May-December 2020: Strategy generates consistent small profits. Portfolio reaches $1,100,000.
The maximum drawdown for this period was 13.73%. This occurred during the initial divergence. Mean reversion strategies are susceptible to trend-following market regimes. These regimes cause prolonged divergence from the mean. This leads to deeper and longer drawdowns.
Another example: a volatility mean reversion strategy. It sells options when implied volatility (VIX) is high and buys when it is low. From January 2018 to December 2018, the "Volmageddon" event occurred. A strategy selling VIX futures when VIX was above 20 might have performed well in calm periods. In late January/early February 2018, VIX spiked from around 12 to over 50. A strategy holding short VIX positions would experience a rapid, severe drawdown. Consider a hypothetical portfolio with $1,000,000 on January 26, 2018. It held short VIX futures. By February 6, 2018, the portfolio value could have dropped to $200,000. This represents an 80% drawdown in a few trading days. This illustrates the extreme risk of mean reversion strategies when the "mean" itself shifts or when deviations become extreme.
Mitigating Drawdown in Mean Reversion
Traders employ several techniques to manage and mitigate drawdowns in mean reversion systems.
Position Sizing: Allocate a fixed percentage of capital per trade. For example, risk no more than 1% of capital on any single trade. If the strategy has a 10% stop loss, then the position size should be 10% of the account value. This prevents single large losses from devastating the portfolio.
Stop-Loss Orders: Implement strict stop-loss levels. For a pairs trade, if the spread widens beyond a predefined threshold (e.g., 3 standard deviations), close the position. This limits potential losses when a pair diverges permanently or for an extended period. For instance, if the average daily true range (ATR) of a stock is $2, a stop-loss at 2x ATR ($4) from the entry price might be set.
Diversification: Apply mean reversion strategies across multiple uncorrelated assets or markets. A portfolio of 10 uncorrelated mean reversion pairs will likely have a lower overall MDD than a single pair. If one pair experiences a large drawdown, others may perform well, cushioning the blow. For example, running a pairs trade on MSFT-AAPL, another on XOM-CVX, and a third on EUR/USD-GBP/USD. Their drawdowns will likely not perfectly coincide.
Dynamic Risk Adjustment: Reduce position sizes during periods of high market volatility or strong trending behavior. If the overall market trend strength indicator (e.g., ADX) rises above a certain level (e.g., 30), reduce exposure to mean reversion trades. This protects capital when market conditions are unfavorable for the strategy.
Look-back Period Optimization: Adjust the look-back period for calculating the mean. A shorter look-back period (e.g., 20 days) makes the mean more responsive to recent price action. This can reduce drawdowns in rapidly changing market conditions. A longer look-back (e.g., 100 days) offers more stability but might react slower to shifts.
Out-of-Sample Testing: Rigorously test the strategy on data not used for development. This reveals potential weaknesses and identifies periods of high drawdown risk. A strategy that performs well in backtesting but shows massive drawdowns in out-of-sample data is unreliable. For example, test a strategy developed on 2010-2015 data on 2016-2020 data. Observe the MDD in the out-of-sample period. If it is significantly higher, re-evaluate the strategy.
Example of Drawdown Mitigation: A pairs trading strategy on SPY-QQQ. Initial capital: $500,000. Backtested MDD (2010-2019): 12%. Live trading starts January 2020. Position sizing: Max 1% risk per trade. Stop-loss: 2.5 standard deviations of the spread. During March 2020 market crash: The SPY-QQQ spread diverges significantly. The strategy generates a stop-loss signal. The trade closes with a 2% portfolio loss. Without the stop-loss, the divergence could have led to a 10% loss on that single trade. The overall portfolio drawdown for March 2020 was 5%. This was lower than the backtested MDD, due to implemented risk controls.
Implementing these mitigation techniques does not eliminate drawdowns. It reduces their severity and frequency. It enhances the long-term viability of mean reversion strategies.