What are possible ML tasks we can achieve with the data model

There exists good, closed form AI to play the game, and we don't exactly want it to just be an AI player. I am thinking the more interesting ML questions are:

## Straight forward

- **Regression**: predict final game score based on current game state
- **Anomaly detection**: was that last piece placement not a good play?
- **Classification**: of what strategy the player was using (would require lots of labeling I think)

## Coach mode

The end goal is to make a coach for the player which could (all options):

- Indicate when to switch strategies based on game state
- lines/tetris
- where to put well
- post game stats
- how many tetrises to goal

The issue lists possible regression, anomaly detection, classification, and coach-mode tasks, but names no files, tests, or entry points. Start by inspecting the repository's data model and existing game-state analysis, then confirm one narrowly scoped task with the maintainer. Done criteria are not defined in the issue and would need to be agreed before implementation.

Written by the indexing model from the issue text.