Bot Found Trades, Never Executed Them

The Scanner Found Trades the Executor Always Rejected

A trading bot failed to execute any opportunities because its execution stage incorrectly calculated available margin, rejecting every trade despite sufficient account equity. This was caused by a divergence in how the scanning and execution stages interpreted account margin, specifically for futures accounts at one broker.

What went wrong

Our automated trading system, Tradewink, was identifying profitable trade setups consistently. The scanning component of our pipeline would flag these opportunities, log them as actionable, and pass them to the execution stage. However, every single one was subsequently rejected by the executor with an "insufficient margin" error. The reported available margin was consistently zero, despite the account holding approximately $350 in real equity. This meant that while the system was brilliant at finding potential trades, it was incapable of acting on them.

Why it happens

The core of the problem lay in how two distinct stages of our trading pipeline computed available margin. The scan stage was designed with a fallback mechanism: it first checked for positive buying power, and if that was zero or negative, it would use the account equity as the available margin. This logic correctly reflected the actual capital available for trading.

The execution stage, however, had a more rigid check. It only considered buying power and cash. Crucially, for futures accounts at a specific broker we were using, both buying power and cash were reported as zero. This broker reported net liquidation value instead of the traditional buying power and cash figures that our execution logic expected. Because the execution stage lacked the equity fallback present in the scan stage, it saw zero available margin for every futures trade, leading to the constant rejections. The scan stage, meanwhile, logged the opportunities normally, making the pipeline appear healthy until the final execution gate.

What we changed

To rectify this, we implemented a unified approach to margin calculation across the entire trading pipeline. The rule we adopted is straightforward: any time an account's buying power or cash is read for a margin check, account equity is now the last-resort fallback. We refactored our code to use a single, shared helper function for all margin checks. This ensures that both the scan and execute stages, and indeed any other part of the system that needs to know available capital, use the exact same logic and data interpretation. This eliminated the discrepancy and allowed the execution stage to correctly identify the available equity when buying power and cash were reported as zero.

How to check your own system

If you suspect a similar issue in your trading bot, perform the following checks:

• Review Margin Calculation Logic: Compare how your system's scanning/signal generation components determine available margin versus how your execution component does. Look for differences, especially in fallback mechanisms or how specific asset classes (like futures) are handled.
• Broker API Data: Verify the exact fields your broker's API returns for buying power, cash, and equity. Understand how your broker reports these values, particularly for different account types (e.g., cash vs. margin, futures vs. equities).
• Test with Known Equity: Manually set up a test account with a small, known amount of equity (e.g., $350 as in our case). Run your bot in a simulated environment and observe the reported available margin at both the signal and execution stages. Does it match your expectation?
• Log Everything: Ensure your system logs the precise values used for margin checks at every relevant step. This data is invaluable for debugging discrepancies.

This systematic approach can help uncover hidden bugs that prevent your bot from acting on opportunities it correctly identifies.

Disclaimer

This article describes engineering decisions in a trading system. It is not investment advice. Trading involves substantial risk of loss and is not suitable for all investors. Past performance does not guarantee future results. Always do your own research and consider your financial situation before trading.