runSimulation method

void runSimulation(MCTSNode root)

Performs a single MCTS simulation (Select -> Expand -> Evaluate -> Backup)

Implementation

void runSimulation(MCTSNode root) {
  MCTSNode current = root;
  List<MCTSNode> searchPath = [current];
  List<int> actionPath = [];

  // 1. SELECT: Follow UCB until we find a leaf or an unexplored action
  while (current.children.isNotEmpty) {
    int action = _selectAction(current);

    if (!current.children.containsKey(action)) {
      actionPath.add(action);
      break;
    }

    actionPath.add(action);
    current = current.children[action]!;
    searchPath.add(current);
  }

  // 2. Handle first simulation or root-only expansion
  if (actionPath.isEmpty) {
    actionPath.add(_selectAction(current));
  }

  // 3. EXPAND & BACKUP: Internal call to the dynamics head and backprop
  _expandAndBackpropagate(searchPath, actionPath);
}