Teacher mode overlays AutoGo's own analysis on the board so you can
see where it would play and why. One visualization is shown at a
time; pick it from the dropdown. Games where teacher mode was ever
enabled do not count toward the leaderboard — we add rank unaided wins to the leaderboard.
- MCTS value:
the search-refined win probability for playing each top move (closer to 1 is better for you).
- Policy priors:
the raw probability the policy network assigns to each of the top moves — good for spotting which stones the network considered without search.
- Raw v_theta:
the value network's win-probability estimate for the position after each top move, before MCTS averaged anything in. Disagreements with MCTS value highlight moves whose "gut instinct" improved or worsened after search.
- Max depth:
how deeply MCTS explored each top move. Low numbers mean the search didn't invest in that branch; high numbers mean it went on a long read.