Introduction
One of the most important challenges in algorithmic trading is controlling losses during drawdowns while maintaining upside potential. A commonly used approach is to reduce position size when the trading system’s equity drops below a predefined threshold (e.g., 10% from peak). When equity recovers above the threshold, position size is restored to full.
While effective during sustained downturns, this method can cause problems when the equity curve oscillates around the threshold. In such cases, the system frequently switches between full and reduced size, often re-risking right before a loss, exacerbating drawdowns instead of protecting capital.
This article investigates this phenomenon and proposes refined methods that preserve the protective benefits of equity thresholds while improving performance in sideways or volatile environments.
The Drawdown Threshold Mechanism
The basic idea is simple:
- When equity < threshold: Reduce position size (e.g., 0.5x)
- When equity > threshold: Restore position size to 1x
This is intended to:
- Reduce exposure during sustained losses
- Preserve capital for future recovery
However, this logic has a serious flaw when the equity curve experiences short-term reversals or noise near the threshold.
Graph: Example of drawdown threshold implementation based on raw price with 0.25 reducer below threshold.
Problem Example:
- Equity drops below the threshold → size reduced
- A slight gain moves equity above the threshold → size increased
- A loss immediately follows → full-sized loss pushes equity back below the threshold
This kind of chop introduces unnecessary drawdown and can erode the system’s overall performance.
Solutions and Enhancements
1. Buffer Zone (Hysteresis)
Graph: Example of drawdown threshold implementation based on a buffer zone (Hysteresis) with 0.25 reducer below threshold.
Definition: Add a buffer around the threshold to prevent frequent toggling.
- Reduce size when equity < (threshold – buffer)
- Restore size when equity > (threshold + buffer)
Benefit: Prevents overreaction to minor fluctuations around the threshold.
Widely used in professional systems for its simplicity and effectiveness.
2. Persistence Requirement (Confirmation Delay)
Definition: Require the equity to remain below/above the threshold for a set number of bars (e.g., 3) before triggering a size change.
Benefit: Filters out noise by only acting on persistent trends. Frequently combined with hysteresis.
3. Smoothed Equity Curve
Graph: Example of drawdown threshold implementation based on a moving average with 0.25 reducer below threshold.
Definition: Apply a moving average (e.g., 5–10 periods) to the equity curve and compare it to the threshold.
Benefit: Filters out short-term equity noise. However, smoothing introduces lag, so it’s best used with additional protections (e.g., a stop-loss).
4. Gradual Scaling Instead of Binary Switch
Definition: Use tiers of drawdown to proportionally reduce position size.
Example:
- Drawdown < 5% → 1.0x
- Drawdown 5%-10% → 0.75x
- Drawdown 10%-15% → 0.5x
- Drawdown > 15% → 0.25x
Benefit: Smooths transitions and avoids hard size jumps. Common in volatility-targeting and risk-parity systems.
5. Momentum Confirmation Before Increasing Size
Definition: Require positive momentum (e.g., recent gains) before restoring full size.
Benefit: Prevents the system from re-risking too early during unstable recovery periods. Often used in trend-following models.
6. Drawdown-Weighted Confidence Models
Definition: Adjust not only size, but also model confidence, based on drawdown severity. For example, weight signals lower when in drawdown.
Benefit: Provides a more flexible and dynamic risk model, especially for ensembles or multi-strategy systems. More complex to implement, but powerful.
Best Practices in Professional Trading
Professional traders and quantitative funds rarely rely on a single rule. Most combine several techniques for robustness. A typical setup might include:
- Hysteresis buffer (±1% around threshold)
- Persistence requirement (e.g., 3-day confirmation)
- Gradual scaling based on drawdown tiers
- Optional momentum confirmation before re-risking
These systems are evaluated against both raw strategy performance and capital allocation behavior to ensure they’re not introducing procyclical risk.
Which Equity Curve Should You Use?
This is a critical design decision:
- Use the raw (unleveraged) strategy equity curve to calculate drawdown and apply threshold logic
- Why? It avoids feedback loops where position size changes distort the measurement of the drawdown itself
- Exception: In live portfolio-level risk management, you may use actual equity to monitor account-level exposure, but for strategy logic, raw equity is preferred
Conclusion
Dynamic position sizing based on drawdown thresholds is a powerful tool—but it must be implemented thoughtfully. Naive switching can do more harm than good in choppy markets. Enhancements like buffer zones, persistence filters, smoothing, and scaling provide significantly more robust behavior.
Most importantly, apply these rules to the raw equity curve to maintain logical consistency.
By simulating these approaches and choosing the right combination for your strategy, you can significantly improve resilience and risk-adjusted returns.