Examples

This page collects the main end-to-end examples for the most common poke-env workflows.

Choose an Example

• poke-env Quickstart: Practical Examples and Snippets: your first local battle and first custom agent; requires a local Showdown server.

• Creating a custom teambuilder: custom team selection and generation; requires a local Showdown server.

• Connecting to Showdown and Challenging Humans: connecting to the official or a custom server and interacting with human players; an account is required for the public server.

• Reinforcement learning with Stable-Baselines3: training with action masking and Stable-Baselines3; requires a local server plus RL dependencies.

• Self-play reinforcement learning with SuperSuit: self-play training with SuperSuit and Stable-Baselines3; requires a local server plus RL dependencies.

• Action Mapping and Strict/Fake Modes: understanding environment actions, legality checks, and masks for custom RL pipelines.

• Strict Battle Tracking: debugging request parsing and battle-state desynchronization issues.

• Saving Battle Replays: automatic replay persistence and explicit HTML replay export.

• poke-env Quickstart: Practical Examples and Snippets
  • Creating Agents and Making Them Battle
    • Creating Built-in Agents
    • Creating a Battle
    • Inspecting the Result
    • Running a Cross-Evaluation
  • Building a Max Damage Player
    • Implementing the MaxDamagePlayer Class
    • Testing the MaxDamagePlayer
  • Setting teams
    • Using a str
    • Dealing with team preview
  • Other Initialization Options for Player Objects
    • Specifying an Avatar
    • Saving Battle Replays
    • Logging
    • Concurrency
    • Pokemon Showdown Timer
• Creating a custom teambuilder
  • Creating a custom Teambuilder
  • Building a team from a set of pokemons
• Connecting to Showdown and Challenging Humans
  • Prerequisites
  • Connecting Your Agent to Showdown
  • Challenging a Human Player
  • Accepting Challenges from Human Players
  • Playing on the Ladder
• Tracking battle observations
  • Defining a snapshot
  • Recording snapshots in choose_move
  • Cleaning up to prevent memory leaks
  • Running the example
• Reinforcement learning with Stable-Baselines3
  • Prerequisites
  • Defining the environment
    • Defining observations
    • Defining rewards
    • Defining the environment class
  • Action masking with Stable-Baselines3
    • Features extractor
    • Masked policy
  • Training
  • Evaluation
• Self-play reinforcement learning with SuperSuit
  • How it differs from the single-agent example
  • Prerequisites
  • Defining the environment
  • Converting with SuperSuit
  • Training
  • Evaluation
• Action Mapping and Strict/Fake Modes
  • Prerequisites
  • Why This Matters
  • Singles Example
  • Doubles Example
  • Recommended Workflow
• Strict Battle Tracking
  • Prerequisites
  • When To Use It
  • Caveats
• Saving Battle Replays
  • Prerequisites
  • Automatic Replay Saving
  • Explicit Replay Export