Quiet failure: the trade that never executed and never got logged

[Fairman]

Quiet failure: the trade that never executed and never got logged

Execution analysis typically focuses on trades that happened. Signals that failed to fill, orders that were rejected, and trades blocked by filters are often missing from the records. That omission biases the picture of strategy performance.

If a system generates signals and half of the winning ones fail to fill while most of the losing ones do, the live results will be much worse than the signal-level results. Without logging the failures, the cause remains hidden.

Log every signal, whether or not it resulted in a trade. Record what happened: filled, rejected, requoted, skipped by filter, expired. Include the time and the reason.

Then compare the hypothetical outcome of all signals with the actual trades. Analyze the unfilled ones by outcome. If missed trades are disproportionately winners, execution has a selection effect that costs money.

This single analysis often reveals the largest leak in a scalping operation.

Are signals that did not fill logged and analyzed in the current setup?

[Amanhasnoname]

This was the biggest blind spot in my own testing. I run a semi automatic setup that fires an alert and I enter by hand, and for months I only recorded the trades I actually took. When I started noting every alert in a spare column, I found that about a fifth of them were missed because I was in another trade or away from the desk.

The surprise came when I compared outcomes. The signals that I skipped because of a busy platform were slightly better than average, partly because they came during fast moves when my attention was already full. The cost was not huge per trade, but over a quarter it came to a few hundred pips.

I now give each alert a status: taken, missed, filtered, or rejected, with a short reason. A weekly count of each status shows quickly where the leak is, and for me it led to a rule that I do not hold two trades open during the London open.