Machine learning crypto trading! Browser-based creation of a crypto bot! Multi-exchange three-way crypto arbitrage!?!?!
Yep — Just want to jot a short note about the future of Bowhead and display some of the progress.
I am still working on Bowhead however that work has been slowed as I have switched jobs and I have also switched the purpose and style of the system. Bowhead was also broken by one of the exchanges which was critical to the primary function of how it was set up and so I needed to rework the entire system.
So many of you have emailed me requesting that I help, or requesting a web configuration that I have started down that path and while doing that I realized that I could move much, if not all, of the creation of your own bot’s into the web interface and then also add in graphs. So, there has been a large effort on my part to do this.
Additionally, I didn’t want the system to be locked into one exchange moving forward as that was clearly an issue so I have made it where there are two options.
- Quickly get set up with Coinigy where you will store your exchange api keys in the cloud
- Manually set up locally with CCXT where you will store your api keys locally and security will be up to you.
Both of these systems allow for data collection, trading and transfer between many different exchanges. (Coinigy has around 14-20 major exchanges it works with and CCXT allows for about 90 different ones)
With this kind of coverage there will be some great ways to add in things like standard arbitrage and three-way arbitrage, which is little different in that it uses three pairs for the arb opportunity.
You can try out a lot of this already with Bowhead, the web configurator is mostly complete and I am working now on the interface.
There is MUCH more I am adding as well..
Bowhead now has support for the Postgres plugin TimescaleDB which is seamless.
Timescale | an open-source time-series SQL database optimized for fast ingest, complex queries and…
An open-source time-series database fully compatible with Postgres for fast ingest and complex queries.
It can be found here, just fire up a new ubuntu AWS ec2 instance or a new Linode and log in and download the raw version of this script from here.
chmod +x SparklingWater.sh && ./SparklingWater.sh
and keep an eye on the output as you will see the URL to your new h2o instance.
This will provide you with a fully functional high-end machine learning system that is ready to go.
I am not quite ready to start dropping what I have been doing with this yet, just suffice to say there is a lot of cool stuff going on with h2o, for instance, just check out the driverless AI they have developed for GPU systems.
See Venkatesh explore how driverless AI is helping PayPal to keep fraudsters at bay and share results from experiments…
Also note that You can use Google colab
Train Your Machine Learning Models on Google’s GPUs for Free — Forever
Training your model is hands down the most time consuming and expensive part of machine learning. Training your model…
And a side project:
Another thing is I have been working on is budget hacking. Most programmers are kind of interested in money because it’s simply just numbers. So, a lot of very creative and interesting individuals have figured out ways to be financially independent and to retire early. I am interested in how this ties into two of my other interests, geo-arbitrage and real estate assets. This is a blog of my research and collected resources for those three topics combined.