Advanced defense: detect alternative paths to same macro square; prefer global-line wins that capture multiple boards

Let players choose AI start order

Add idle self-play training

Precise owned-board penalty: only penalize if move didn't win a new board via global line

Null owned cells in heuristic global threat analysis; bump owned-board penalty to -20

Swap priority: block opponent micro-board threats before taking speculative wins

Add board diversity bonus to heuristic and opening spread incentive

Upgrade from 2-ply to 3-ply NN minimax with wider search (TOP_K=14, OPP_TOP_K=12, AI_DEPTH3_K=8)

Run as root (HF Spaces default) so /data storage bucket is writable

Try /data storage bucket with runtime fallback to ./ai/ if write fails

Drop bucket mount approach, persist via Hub API pushes on each game save instead

Seed model weights to DATA_DIR from git before training (fixes ENOENT /data/model_weights.json)

Fix bucket write permission: entrypoint runs as root, chmod 777 /data, then drops to appuser via su-exec

Fall back to ./ai/ if /data is not writable (EACCES on mounted buckets)

Add logging for DATA_DIR and game saves to debug recording

Persist game data and trained weights via HF Hub API (no bucket needed, uses built-in HF_TOKEN) + defer server start until init complete

Fix loadGames is not defined (missing import after fork refactor)

Auto-detect HF /data persistent storage; no env var needed

Store training data and trained weights in DATA_DIR (configurable, survives HF Space restarts); fall back to git-tracked model_weights.json

Run AI training in child process (fork) so server stays responsive during training

Auto-train AI every 5 games: background training, games-since-last-train counter, UI shows pending status

Train AI from human games: game recording, JS backprop training, value head init, Train AI button, record toggle

Replace heuristic 3-ply with NN-based 2-ply minimax + fix game-level fork detection

Fix AI tactical priorities: offence before defence, fork detection, global lines, avoid owned boards

Rename 'vs AI' to 'AI' on home screen card

Fix 'Play Again' for AI mode — reset local board state instead of emitting restartGame (which triggered wrong online flow)

Fix AI: heuristic compared numbers to strings (never detected threats). Replace MCTS with 3-ply minimax + explicit opponent board-win blocking. Increase board ownership weight from 10 to 20.

Improved heuristic (near-wins, threats, cell counts) + 150 MCTS sims for tactical AI

Export value_head in model for neural network value evaluation

Add MCTS at inference: 50 sims, heuristic+network eval for strong tactical play

Update trained model weights (AlphaZero, 244K params, 3 gen)

Update resume path for Colab, reduce gens/games

Fix Dockerfile: copy ai/ directory for AI engine

Add AI mode: neural network opponent with MCTS-guided self-play training

Adjust chat badge position: higher (-6px) and more to the left (right: 0)

Fix chat badge cut-off: use inner wrapper for badge positioning instead of overriding fixed position on toggle button

Merge branch 'main' of https://huggingface.co/spaces/sebebeb/supertictactoe

Mobile UX improvements: touch-safe hover effects, tap delay removal, larger board cells, unread chat badge, compact layouts for small screens

Update server.js

Full-page background tint matches current player's color

Background tints to current player's color; reset on home screen

End game when all macro squares are owned (majority wins, tie=draw)

Fix: lines through owned boards still count; only award unowned boards in the line

Revert: owned board cells are NOT nulled in global line detection grid

Remove board-sending mechanic, restore free-play; keep owned-board cell nulling in global line detection

Fix: null out owned board cells in global line detection; add board-sending mechanic

Update public/js/app.js

Update server.js

Upload 9 files

initial commit