What happens when a trading bot loses money?

Whose loss is it, and what bounds it

Buying trading software is buying a tool: the vendor is paid for the tool, and market outcomes — both directions — belong to the operator. Any arrangement claiming the vendor absorbs your downside is either a regulated product wearing a costume or a fraud building trust before an exit.

What you control is the bound. An isolated sub-account caps the absolute worst case at the compartment balance. Position sizing caps each individual mistake — a desk trading $5 fractional clips cannot have a catastrophic single trade. And a kill switch caps duration: when behavior looks wrong, you stop it now, not after support answers.

• Vendor liability: none, in any honest software arrangement — plan accordingly.
• Absolute bound: the isolated compartment's balance.
• Per-trade bound: position sizing rules you can read before running.
• Duration bound: your kill switch.

A losing day on a well-run desk

On a desk with a real journal, a losing day is a legible event: which position, on which signal, under which rule, closed by which decision. You can trace it, judge whether the process was followed, and decide whether the rules still deserve to run. That is what 'losses included' means on our public tape — losing entries print in the same font as everything else.

The alternative is the mysterious loss: a balance that dropped for reasons the product cannot explain. Treat explainability as a purchase criterion — a bot that cannot show its reasoning on a losing day was never showing you its reasoning at all.

First-party data — from our own desk

Keep going