So, after observing my secret project in motion for a while and determining it to be largely a failure, though one with interesting and odd results that show some promise, and after spending a few weeks mulling over what the potential impact would be if I published what it was, I’ve decided to do just that.
Tetsuo is a closed source stock trading system that I’m working on, which I will never release to the public.
It downloads stock historical data, performs a few proprietary regression analytics, and performs time series forecasts to try to determine which stocks will go up or down, and buys the winners.
The first version was a weekly trader, which, never deviated from 2% of its entry, but broke even after a few months.
The version I’m currently working on, or rather have been complaining about not working on due to time constraints, is a daily trader with some improved approaches. It seems to be showing more profitable results, but, time will tell. Gotta get those NRMSE, SMAPE values down.
At least now I can talk about it without having some ominous “secret project” being referred to, because it occurred to me that sounds weird.
The trader is called TETSUO. I’m actually rather proud of some pieces of it, so it’s a pity I can’t share, as it culminates a great deal of time spent starting at an IDE.
In terms of performance, it’s not quite ready for putting in action, as, I’ll get 10 good days and a really bad one, and, with the NRMSE and SMAPE values being a little higher than I like as well, accuracy and consistency are concerns.
My current approach at selection criteria is flawed as are my forecast quality metrics calculations currently. For selection criteria I’m using forecast metrics as filters, which should be okay for filtering out outliers, but I’m unable to find good balanced values for any of the metrics, and, I’m finding that the correlation between quality metrics and values is overall lower than I aimed for.
So, it needs another round of development. When I’ve not been crawling away from my desk for a while.
This time barrier is ridiculous but there’s a light at the end of the tunnel coming.