🔮 PurrDiction

PurrDiction is PurrNet's client-side prediction addon. It lets the controlling client simulate input immediately while the server runs the authoritative simulation. When a verified server frame arrives, the client restores the verified state, replays newer inputs, and smooths any visible correction.

Use PurrDiction for gameplay where waiting for a round trip would feel bad: player movement, vehicles, projectiles, physics interactions, combat state, or other tick-based systems.

The mental model#

Every predicted object has two layers:

• Simulation runs at the network tick rate. Inputs and state are saved so PurrDiction can roll back and replay them.
• View runs once per rendered frame. It turns corrected simulation state into smooth visuals without changing gameplay.

The server remains authoritative. Clients send input, not trusted gameplay state.

flowchart LR
  A["Read local input"] --> B["Predict immediately"]
  B --> C["Send input to server"]
  C --> D["Server simulates"]
  D --> E["Verified frame returns"]
  E --> F["Rollback and replay"]
  F --> G["Smooth the view"]

Start here#

1. Install and configure PurrDiction.
2. Read the architecture overview.
3. Choose a prediction policy for each kind of object.
4. Implement a Predicted Identity, or begin with a built-in component.
5. Review input handling, state handling, and the execution flow before building more complex interactions.