Automation, can we?


#1

I know in the past automating uploads was a huge problem. Not sure where that stands. Would be cool if we had an api or a process for downloading and uploading right from python, would streamline the process. Maybe we can and I just haven’t figured it out, would like to know the official stance before I dig too deep. Don’t need my accounts getting whacked.


#2

There’s an unofficial API at https://github.com/numerai/NumerAPI which I think still works for some stuff at least, though I haven’t used it to confirm.


#3

I use the unofficial API to upload my predictions and it seems to work fine most of the time. I sometimes get no feedback from the python interface on whether a submission was successful or not, but it works well enough.

I’ve not used any of the other functionality of the API though.


#4

Thank you both. I’ll play with this a little see what I can do.


#5

What about an API for utilizing those predictions in the market? Does something like that exist, or is going to exist soon?


#6

Simon I can give you my narrow perspective on this. As far as a direct API to the broader market, I don’t know, but i doubt it. The “optimized model” you create here won’t be useful because we don’t know the exact underlying data. However, the process, the framework, you develop here is absolutely meaningful. That’s based on having used Genetic Programming both professionally and personally for almost 20 years in FX.

With that said, being “right” about a direction or a target is completely different than surviving market volatility and other nearly infinite risks. So if you do decide to take your approach live elsewhere, do a lot of research before hand and test very, very small allocations, for an extended period of time. Being “live” is completely different than doing it on paper. Time slows to a crawl and every single move you dismissed as meaningless while prepping will feel like a complete meltdown when your money is on the line.