ANALYSTICS PAGE MORE GRAPHS

Fix auto-training and add dynamic MVP/Championship predictions

Update continuous learner and git attributes

Fix: Confidence meter shows model confidence, reduced API timeout

Fix confidence meter percentage display and restore badge

Add analytics, confidence meter, enhanced H2H, daily MVP refresh

Add pre-cached 2025-26 starters fallback for HF deployment

Fix nbainjuries import to handle missing JVM gracefully

Fix auto-training, injury-aware lineups, and nbainjuries API

Fix: Move warm_starter_cache function before startup code calls it

Fix: Add retry logic and longer timeout for NBA API to ensure real data on HF

Fix: Use pts field instead of ppg in fallback data to match frontend

Fix: Synchronous cache warming for accurate starters and PPG data

Perf: Lightning-fast caching - startup warming, cached rosters, fast fallbacks

Fix: Use real NBA API for starting 5 lineups instead of hardcoded data

Add data/api_data and models with LFS tracking

Fix: Include data/api_data in Docker build

Fix: Add data/api_data to Dockerfile COPY for ELO and standings data

Fix: Add git lfs pull on startup to ensure data files are available

Debug: Add logging to diagnose api_data path issue on HF Spaces

Fix: Copy api_data to persistent storage and use env vars for paths

Add essential data files for production deployment - Includes standings, games, teams parquet files - Fixes predictions showing 54/47 on Hugging Face

Fix: Enhanced prediction algorithm for accurate, varied probabilities - Uses multi-factor Log5 formula with current season standings (40%), recent form (30%), ELO (20%), injury impact, and home court advantage - Replaces fixed 64/36% predictions with realistic varied percentages

Add onnxruntime + tokenizers + caching for fast loading

Smart training after games complete + MVP retry with 60s timeout

Add full features: auto-training, persistent storage, background scheduler

Fix Docker build: use npm install and include package-lock.json

Use environment variables for ChromaDB credentials

Initial commit: NBA Sage Predictor for Hugging Face Spaces (with LFS for large files)