Choosing a gold trading bot: the criteria that matter

Why "best" is the wrong question

Any list ranking gold bots is ranking marketing, because there is no public standard by which they can be compared. Track records are self-reported, backtests are self-configured, and the metric most heavily advertised — win rate — is meaningless without the payoff ratio beside it.

The productive question is not which is best but which has been measured. That is answerable from published evidence, and most products fail it immediately.

The five questions, and our answers

[Maximum drawdown](/learn/glossary/max-drawdown): roughly 21.2% (Jan–Jul 2026) and 20.5% (Aug–Dec 2025). Both around a fifth of the account.

[Sample size](/learn/glossary/sample-size): roughly ten trades a month, which is small enough that no percentage here should be treated as precise.

Out-of-sample: the Aug–Dec 2025 period was never used in development. It returned +$154 and was positive in three months of five — profitability generalised, "green every month" did not.

[Stop loss](/learn/glossary/stop-loss): server-side, attached immediately after each fill. If the modification is rejected the position is force-closed rather than run unprotected.

Failures: 34 strategies tested, 4 enabled. The other 24 are published with their numbers, including the one that looked best in replay and lost money live.

What no bot can do

None of this makes a bot safe. A 20% drawdown is a large number, gold is a volatile instrument, and the broker minimum lot can push real risk above whatever you configured on a small account. Read the risk disclosure before deciding what you can tolerate.