Risk & Drawdowns

Risk & Drawdowns

Risk and drawdowns are central to understanding how an AI agent behaves over time.

Returns alone do not describe how a strategy operates. Risk metrics help explain the variability of outcomes, the severity of losses, and how an agent responds during adverse market conditions.

What drawdowns represent

A drawdown measures the decline from a previous peak in performance to a subsequent low.

Drawdowns show:

• The maximum loss experienced during a period

• How deep losses became before recovery

• How frequently losses occurred

Every strategy experiences drawdowns. The size, duration, and frequency of drawdowns are often more important than returns.

Types of drawdowns

When reviewing drawdowns, consider:

• Maximum drawdown: the largest peak-to-trough decline

• Duration of drawdowns: how long it took to recover

• Frequency: how often drawdowns occurred

Short, shallow drawdowns indicate different behavior than long, deep ones.

Volatility and variability

Volatility describes how much results fluctuate over time.

Higher volatility:

• May produce larger short-term gains

• Often comes with larger or more frequent losses

Lower volatility:

• May produce smoother results

• Often grows more slowly

Volatility should be evaluated together with drawdowns and returns.

Risk in backtests vs live execution

Backtested risk metrics are based on historical data and simplified assumptions.

Live execution may introduce additional risk due to:

• Market liquidity changes

• Slippage and execution delays

• Fees and funding effects

• Capital constraints

As a result, live drawdowns may differ from historical simulations.

Using risk metrics responsibly

Risk metrics help you understand behavior, not avoid losses entirely.

When evaluating risk:

• Focus on whether drawdowns are acceptable for you

• Avoid comparing agents using returns alone

• Consider how the strategy behaves during unfavorable conditions

No strategy is risk-free.

Responsibility and expectations

Risk metrics do not eliminate uncertainty.

Users remain responsible for:

• Choosing strategies that match their risk tolerance

• Monitoring performance over time

• Managing capital and exposure

AITA provides transparency into risk characteristics, not protection from loss.