Home▸AI & Machine Learning▸3D Monte Carlo Tree Search

🎯 3D Monte Carlo Tree Search

3D Monte Carlo Tree Search – the same UCB1 selection, expansion, rollout, and backpropagation as the 2D original, grown as a genuine spiralling 3D branch structure you can orbit and tune live. MySimulator.uk

AI & Machine Learning3DModerate60 FPS📱 Mobile-adapted

⚙ Under the hood

Selection walks the tree using UCB1 — winRate + C·√(ln(parentVisits)/childVisits) — to balance exploiting known-good moves against exploring under-visited ones; expansion adds one child node for an untried move; rollout plays uniformly random moves to a terminal state; backpropagation updates visit and win counts back to the root. This companion runs the identical statistics as the 2D original, but lays the tree out in genuine (x, y, z) space: each branch's direction is rotated from its parent by a fixed pitch/yaw fan per move plus a golden-angle twist per depth, so the tree spirals outward through volume instead of collapsing onto a flat plane.

Monte Carlo Tree SearchUCB1Backpropagation

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install