I’m considering experimenting with some long term personal investment strategy by coding it. I have a hard time approaching investments from a manual research standpoint when I approach everything else from a data + formula standpoint, so I’m looking at the Ally Invest API to try to craft something.
Initially (first year or so) I’m not planning on actually executing any trades, just gathering information and letting it simulate actions to see how it performs.
Since there’s such a variety of experience on this board, I’m wondering if anybody else has done something similar and if there were any suggestions. Just for my own personal Elixir enjoyment reasons, I’m thinking of using GenStage to ingest and consume the investment data, compressing it into Postgres using the CStore extension and then alerting me with recommendations with the formula has something it thinks I should act on.
Any feedback or gotchas would be appreciated.
Here are a few trading-related Elixir projects on GitHub you might find useful (I haven’t tried any of them).
elixir_trading_platform - A realtime trading platform powered by phoenix and websockets.
lgthorn - Cryptocurrency trading bot.
tai - A trading toolkit built with Elixir that runs on the Erlang virtual machine.
talib - Technical Analysis Library written in Elixir.
I’m currently writing an automated trading tool against alpaca.markets. Unless you’re doing crypto, you probably won’t need the bells and whistles of GenStage. I’m wrapping an API endpoint in an HTTPoison client and a WebSocket client in WebSockex and feeding through to GenServer procs. Biggest hurdle ATM the moment is designing a resilient system when the system of record is external, i.e. the website. I.e., re-syncing on failure. Happy to share any hard lessons I’ve learned or gotchas in more detail.
Thanks! Prefer PM or here?
GenStage is complete overkill and only intended for personal tinkering sake.
Either’s fine with me.
Re: GenStage, totally understand. I had the same inclination coming from data pipelines and Kafka. I’m having to re-wire my brain to adopt the let it fail and pick up the pieces mentality.
Currently on my third revision of the codeline, but I feel like I’m getting closer to the best Elixir/OTP way to do it. I.e. more, smaller procs and using bounded contexts and pure functions
I’ve thought of doing similar myself (and/or do some side projects for pricing predictions). For work we’re using rocksdb for IoT timeseries currently and syncing between devices and it’s pretty straightforward to get a basic streaming web hooks going. You might be well served with a KV store as well using ranges and streams, depending on your analysis type.
It’d be interesting to hear if you do anything on this front. Best of luck!
Good idea! I wasn’t even thinking about LiveView but that makes good sense.