---
comments: true
date: '2024-04-12'
description: search algorithm using random sampling, including tree search with selection, expansion, simulation, and backpropagation phases.
id: Monte-Carlo
modified: 2026-06-05 15:08:28 GMT-04:00
tags:
  - seed
title: Monte-Carlo methods
created: '2024-04-12'
published: '2024-04-12'
pageLayout: default
slug: thoughts/Monte-Carlo
permalink: https://aarnphm.xyz/thoughts/Monte-Carlo.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
## tree search.

a [[thoughts/Search|search]] algorithm based on random sampling of the search space.

- Selection: root $R$ and select successive child nodes until leaf $L$ is reached.
  - The root is current game state and leaf is any node that has a potential child from no simulation
- Expansion: Unless $L$ ends the game decisively for either player, then create one (or more) child nodes and choose node $C$ from one of them.
- Simulation: Complete **one** random playout from node $C$.
- Backpropgation: Result of playout to update information in nodes on path from $C$ to $R$.

## simulations