Building a dev tool or hiring developers? Get seen here. Become our Sponsor
Next: let the forward test run, and see if the edge holds up live.
I built PulseShift to answer one question: does my AI trading strategy actually work? At first, no. It lost money after fees in all 16 markets I tested.
So I rebuilt it around one rule: only trust results on data the model never saw.
Separate AI models per timeframe, because fast candles snap back while daily candles trend
Tested 6 published trading ideas; only trend following survived (Sharpe 1.35 vs 0.77 for buy-and-hold, worst drop −22% vs −76%)
Buy, exit and target signals drawn automatically on the chart
A 180-day forward test to check the edge holds up live
Lesson: most "AI trading" edges vanish once you count fees and test on new data.
Building a dev tool or hiring developers? Get seen here. Become our Sponsor