Hi everyone,
While I am really impressed with the data portfolio123 provides, I’ve noticed that many stocks jump right after earnings releases, before the new fundamentals show up in Portfolio123 via FactSet’s updates. As I understand this is due to real-time parsing.
Has anyone tried integrating real-time earnings parsing or alert systems into their P123 workflow? I’m curious about tools or methods that let you flag big data surprises early & systematically and feed a temporary “pre-FactSet” signal into your ranking model. Any tips or experiences would be much appreciated!
Yes, this is a real thing. I haven't done it but know it's been done for many years. I'd recommend using one of the LLMs via API. Somewhat of a lift but that's probably the cheapest way. It may be easier to trade post earnings drift - though far less profitable with more risk.
1 Like
This is my daily task.
I don't yet have a very advanced tool, but what I have is sufficient to make a reasonably good guess.
Before the market opens, I call the model gemini-2.5-flash-preview-04-17 via the Python API and ask it to retrieve all relevant financials from the latest Form 10-Q. I’ve found gemini-2.5-flash to be the most accurate for this task.
Once I retrieve the financials, I calculate ratios and create a pseudo-model that more or less mimics my actual ranking system in P123. Then, I estimate the probability that a stock will fall below the RankPos threshold. I had to train this model earlier.
In any case, everything is still at a very early stage.
3 Likes