I’m going to necro this thread because it honestly is a fantastic freeze-frame of the things we were thinking about “way back when”. For those that haven’t been following over the past many months I’ve been building out Mimir in totally random in line with my “manifesto” found here:
Mimir development has hit a point where it’s time to start thinking about exactly the things this post puts forward. Mimir is loaded with news headlines, earnings data, economic data, fundamentals, quotes and more now and it’s all in a system that makes it easy to utilize functions written in any language. For those that haven’t seen it, I’ve released a development playground for Mimir and will continue to build upon the available connectors as interest increases. For more on that you can read here: GitHub - AscendedTrading/Mimir: The system that powers Valhalla and all it's features
What we learned in the early stages of building out the community’s tech is that if we just build bots that dump into channels the information goes unused. So we needed a full system capable of not just aggregating information but one that could actively make decisions based on that information and the foundation for that is now there.
Current ideas that I’m aggressively going to be attacking are building out Mimir’s ability to gauge sentiment of news headlines in two distinct ways.
I’m going to use basic models to calculate and display “How the algos are probably reading it”. A display designed not to be accurate in line with how we’ve observed in the market but instead to represent the “baseline”.
I want to work on our own in-house model designed to ingest and score headlines as they come in, initially based on user input (Buybacks = 10 points importance, CEO departure = 10 points, Winning an award = 5 points) but to also track stock performance of those headlines after the fact in an attempt to generate a process that can watch the stream of headlines and select ones that are most likely to have a substantial effect on the underlying.
Another area that I’d like to put heavy development into is earnings. Mimir now has an earnings database and can calculate percentage beats, with some criteria added to that, we can generate signals in the morning for what specific earnings plays to pay attention to. This can be using fundamentals data to bolster the results or by scanning SEC filings, a functionality I’ve almost completed.
To be clear, this thread is to discussing creating processes in Mimir to help identify plays. So any idea is welcome.