Hi Kristof,

thank you for your input. Here's my take on those two points.

1) Ideally, I should. It's just a matter of time I guess haha. Sorry...

2) I share your observation that earnings probably have the largest impact on short-term changes. Same for larger events and news (Covid, Ukraine, etc).

Such big news are unpredictable and to integrate more regular news, we need to use language models and larger data pipelines. It's a bit much for my experiments at this point so I left it out. This also includes sentiment, confidence, etc.

But one important note on earnings: They happen four times a year maximum while tick data can be by the minute. This is a challenge when modeling since the largest granularity dictates the overall granularity. Meaning: my input dataframe suddenly has only four rows (for four Qs).

Btw, I also doubt I will get well-performing models, in general. But that's what my post series tries to explore.

Best,

Jonas