basic approach

[jeremymanning]

This is a starter repository. The approach is:

1. create a general purpose package for scraping maps, logs, and conversations from dominating12 (https://dominating12.com/), focusing on a specific set of usernames (to be defined) and also scraping from the top 100 list here: https://dominating12.com/userlist
2. also scrape maps here: https://dominating12.com/maps. turn each into a high quality vector (plottable at any resolution) with labeled territories and connections. each map gets a spatial representation (the map) + a graph representation (which territories can attack?)
3. match up logs and conversations
4. define "position strength" metrics
5. define graph metrics per map
6. look at:

• per map game dynamics: do maps cluster? do games cluster? are there general trends?
• per player dynamics: do players cluster?
• per-game dynamics: do conversations (embed each conversation turn using EmbeddingGemma) follow stereotyped trajectories? are there patterns of what people say as a function of their current position strength, position in the game, and/or position of other players?

7. also create playback tools for displaying animations of game and conversation dynamics