SONIQE
Due diligence guide

How to Evaluate an AI Trading Bot

Ignore the projected monthly return for a moment. A useful evaluation starts with evidence, downside and who controls the money.

1. Live or backtested?

Backtests can be useful for research but are not equivalent to live execution. Prefer a sufficiently long live-account record with a meaningful number of trades.

2. Look at drawdown

Return without drawdown gives an incomplete picture. Drawdown shows how far an account fell from a previous peak during the measured period. Historical drawdown is not a guaranteed future maximum.

3. Understand leverage and amplification

Determine whether the product uses ordinary broker leverage, a separate capital-amplification programme, or both. They are not interchangeable concepts.

4. Calculate net return after fees

Performance fees reduce profitable periods. Model net outcomes rather than repeating gross strategy returns.

5. Follow the money

Identify the broker or custodian, whose name the account is in, how deposits are made and what the withdrawal process is. Strategy quality and custody risk should be evaluated separately.

6. Stress-test the downside

Ask what happens if the strategy loses 0.5%, 1% or more. High amplification can make seemingly small strategy moves significant relative to contributed capital.

Stress-test your assumptions →