What I log so that a bad day is debuggable later

When something goes wrong, "the strategy lost money" is not a diagnosis. Here's what I keep so I can actually reconstruct a session: 1) Every signal that fired, including ones that didn't become orders and why. 2) Intended price vs actual fill, per order. 3) The exact strategy version running, not "the momentum one." 4) Connection state changes with timestamps. 5) Anything I changed manually mid-session, with the time I changed it. Number 5 is the one people skip and it's the one that explains the mystery days.
BacktestBetty
16 posts · 0 followers
+ Follow

Related reading

4 replies

DataDrivenDee· Jul 2026 ago
Number 5 is the difference between a log and an alibi. Most people keep the alibi.
D
DrawdownDave· Jul 2026 ago
I once spent a weekend blaming the software for a day I had personally changed the size on. The log would have told me in four seconds.
H
HalfKelly· Jul 2026 ago
Add one more field: the sizing inputs at the moment the order was built, not just the resulting size. If size = fraction x equity / risk-per-unit, then any of those three moving will change the trade, and later you can't tell which one did it. I log equity, the risk estimate used, and the fraction, and the number of times the risk estimate was stale has been humbling.
C
CoffeeAndCharts· Jul 2026 ago
Number 5 lives in my morning notes now, one line, timestamp and what I poked. Half the time it's nothing, but the days it's something are exactly the days I'd have argued with myself about it a week later. Betty, do you keep the manual changes in the same log as the machine stuff or a separate file?
Sign in to reply →
← All automated trading