Spent three weeks debugging a rule that was never broken

My setup is dull on purpose - 5m chart, two windows, entry only after a full candle closes past the level. Except my entries kept firing early and I assumed something in my conditions was misfiring, so I rebuilt the whole thing twice, rewrote the trigger, kept a log of every mismatch. Turns out I'd been eyeballing the 5m but had a 1m window open on the left and reading off that one half the time. Rule was fine. My eyes weren't. Three weeks of evenings to find out I needed to move a window.
F
FiveMinFiona
14 posts · 0 followers
+ Follow

Related reading

2 replies

BacktestBetty· Sep 2026 ago
Classic. Same family of error as trusting a backtest that silently ran on the wrong timeframe series - I lost a month once because my data feed and my chart disagreed about session start and I kept tuning parameters to fix what was really a clock problem. Two things save me now: I label every window and every dataset with the timeframe in big ugly text, and before I touch a rule I check whether the rule ever actually fired wrong or whether I just think it did. Most of my 'broken strategies' were me overfitting to a misread.
DataDrivenDee· Sep 2026 ago
Is there a way to log what you were actually looking at when you entered, or is that just journal discipline?
Sign in to reply →
← All feature requests