README.md

ArcaThread Processor

Pre-computes champion statistics from matchup-matrix data for fast tier list generation.

What it does

1. Scans matchup-matrix for latest patches every hour
2. Aggregates data per champion (win rates, matchups, by role/rank)
3. Generates JSON files at champ-stats/{patch}/{champion_id}.json
4. Creates tier list at champ-stats/{patch}/tier-list.json

Output Structure

champ-stats/
├── 16.4/
│   ├── meta.json          # Patch metadata
│   ├── tier-list.json     # Sorted tier list
│   ├── 266.json           # Aatrox stats
│   ├── 103.json           # Ahri stats
│   └── ...
├── 16.3/
│   └── ...

Champion Stats Format

{
  "champion_id": 266,
  "total_games": 15420,
  "win_rate": 0.5234,
  "by_role": {
    "TOP": {"games": 12000, "win_rate": 0.5240},
    "JUNGLE": {"games": 3420, "win_rate": 0.5210}
  },
  "by_rank": {
    "DIAMOND": {"games": 3000, "win_rate": 0.5100},
    "MASTER": {"games": 1500, "win_rate": 0.5050}
  },
  "matchups": [
    {"enemy_champion_id": 54, "games": 500, "win_rate": 0.4800},
    ...
  ]
}

Environment Variables

Variable	Default	Description
HF_TOKEN	(required)	HuggingFace API token
DATASET_REPO	ArcaThread/arca-thread-priors	Target dataset
PROCESS_INTERVAL_SECONDS	3600	Run interval (1 hour)
MIN_SAMPLE_SIZE	100	Minimum games for inclusion

API Endpoints

• GET / - Basic info
• GET /health - Status and stats
• GET /trigger - Manually trigger processing
• GET /patch/{patch} - Patch processing status

Deployment

Deploy to HuggingFace Spaces as a Docker space. Set HF_TOKEN as a secret.