I m not clear what kind of

I think so! Would love for you try it out. You could dump statistics for each player as raw text documents into GPT index, and then ask it these questions. Would love to hear your feedback if this is correct!

We're also working on adding structured data support to GPT Index which should help with your use case. In the meantime you could try feeding in text data into GPTSimpleVectorIndex (we'll embed each chunk)

Could this next line be considered a "raw text document" or "structured data"?

Player 2 Pt% 3 Pt% FT% Rebounds
Jones 35.0 15.2 89.0 6.0
Smith 24.5 11.6 78.3 4.2
James 43.3 28.5 92.1 8.4

I think it does pretty well with markdown tables.

Text? "Jones shot 35.0 % from two point range, 15.2% from three point range and 89% for Free throws. Smith made 24.5% for two pointers, 11.65 from three point range, 78% for freethrows, and had 4.2 Rebounds per game" - Would these work equally well, or would data need to be summarized in text chunks FIRST?

Wow, I see you can convert CSV files to markdown tables: https://www.makeuseof.com/tag/create-markdown-table/#:~:text=A%20table%20in%20Markdown%20consists,colon%20(%3A)%20aligns%20cell%20contents.

arminta7 RE: markdown, when I copy hyperlinks in Markdown into a MS Word document, they don't work. Do you know a fix for this?