User made resources



Exploratory Notebooks:

Blog Posts

Example Models For Current Dataset

Thanks for sharing these. As someone with little programming experience (read: basically none), it’s really helpful to have some inspiration for getting started.

Do you have any suggestions for data science or machine learning reading that could help? Honestly, even podcasts or YouTube videos would be much appreciated.



I’ve spent a lot of $$ over the years on programming and machine learning books. I think a really good place to start is with Jason’s “Machine Learning Mastery”. It’s more of a math-lite, top down approach. If you work through his series you’ll find digging into some of the more traditional Python, R and Machine Learning text to be easier to understand. It will also give you a very solid foundation to work from if you choose not to pursue any additional learning. ~ JP


@ObjecScience Thank you SO much! This is exactly the kind of thing I was looking for! I’m all about teaching myself new skills and learning new topics, but without a formal education, finding direction can be difficult. I really appreciate you passing this along, I’ll be sure to let you know how it goes!

pinned globally #5


Thats exactly what i´m asking myself now, what i need to study in order to be able to pragram my own artificial intelligence to play in the markets for me?


4 there are just a ton of resources out there that will get you started. I’ve played with both R and Python and keep finding myself coming back to python, pretty much exclusively at this point. I’m sort of a non-traditional learner and get bored stupid fast and couldn’t hold a train of thought if I was pinned to the tracks, so to keep my interest I HAVE to hack away at code typically something written well above my abilities… Those little discoveries and aha moments are what really drive me forward.

If you want a gentle introduction the books series above will get you up and running quick. It’s also really well supported, Jason is really active on his blog answering questions etc and he’s just genuinely a nice guy. I like supporting people like that.

With the basics under your belt you can follow your own crooked path to discovery. More books, blogs, Github, Youtube and StackExchange will uncover all kind of stuff for you. It’s an amazing adventure and because of the nature of this competition, it will allow you to continually experiment and improve without having to adopt brand new thinking every few weeks. Discover, tune, refine, repeat, all in a consistent, competitive environment that allows for it.

One of the things that has really helped me in the last six months or so is setting up environments. Start with something like Anaconda and then struggle through installing things like Xgboost, LightGBM, Tensorflow, Tensorflo-GPU and LightGBM-GPU. Trust me, when you get done installing Boost on windows (required for LightGBM GPU) you’ll feel like a Rock Star (and you’ll be able to curse in half a dozen different langues.) Working on that lower level stuff, banging it out on a command prompt will give you a lot of insight.

Good luck.


Adding my two tutorials:

A simple NN in PyTorch:

A simple CNN in PyTorch:


Sharing the model I’ve used for quite a while. It wasn’t performing good enough to earn lately, too many better models pushing it down while I didn’t work on it. :slight_smile: It did earn some before (total of $30 USD and ~140 NMR iirc). Not changed for the new format but might help someone else get into numerai. Using mostly sklearn and xgboost with 2-level stacking:


I wrote a Keras callback that prints out your models consistency per epoch and also saves the consistency to the training history. The package is in PyPi so installing it is as easy at pip install numeraicb. You can find a full example here