Reading your bot's order log like an auditor

Most traders review a bot by looking at its P&L. Over a small number of trades, P&L is mostly noise — it tells you very little about whether the configuration is sound. The order log tells you something you can act on immediately: did the system behave as specified?

Four checks worth doing weekly

• Entry fidelity — did every entry correspond to a condition in the configuration, and did every qualifying condition produce an entry? Missing entries usually point to insufficient balance, exchange rejections or connectivity gaps.
• Slippage — compare intended level against actual fill. A consistent adverse gap is a sizing or liquidity problem, not bad luck.
• Exit attribution — how many positions closed on target, on stop, on trail, and how many were closed manually? A high manual count means the automation is not the system you are actually running.
• Rejections and retries — every rejected order is a decision the system wanted to make and could not.

What the log cannot tell you

Fidelity is necessary and not sufficient. A perfectly executed poor strategy produces a spotless log and a declining balance. Reviewing execution weekly and reviewing strategy quarterly keeps the two questions from contaminating each other — and keeps you from tearing up a sound configuration after an ordinary losing week.