Better, but still nothing really notable. That line says it all.
It made “a little” money.
So, while it’s been emulating a roller coaster, I’ve been reading and reading, and tinkering and tinkering, and it might be time to advance away from the vulcan build of MK-IV.
A big breakthrough was using fourier transforms to detect seasonality orders. That’s now in MK-IV locally. MK-IV will abandon prices entirely and focus purely on percent return, since that’s really what the system is geared for. This should also provide better directional accuracy of forecasts as well as enhance data stationarity to improve accuracy of those forecasts.
Unfortunately that’s too large a pivot to say in the same subproject, Vulcan, because some of my metric calculations rely on price forecasts, and I’ll have to redo some of the workflows, which means there’s more restructuring to do.
I’ll need a roman subproject name for this upcoming variant since it’s very metricy, like “Centurion”. So, for now, what’s likely coming is Tetsuo MK-V: Centurion. I don’t know that it’s enough change to change the model version. I guess it doesn’t really matter.
1 step forward, 2 steps back. Gotta get money for the monkey on my back.