loss datasets

AlphaZero Training Metrics: Policy and Value Loss Over 13 Iterations

Reinforcement learning training run tracking policy/value losses, game length, MCTS simulations, and value calibration across 13 self-play iterations.