Name: Model Deployment Lab
Author: MySimulator

Challenger model

True error-rate delta vs champion: +1.5 pp Auto-rollback threshold: 4.0% error Auto-rollback enabled

Rollout control

Current stage: 5% canary

Live stats

Champion error rate—

Challenger error rate—

Requests served0

Each tick fires 500 synthetic requests split across the current traffic stage; the challenger's true error rate is hidden — only its rolling observed error rate is visible, exactly like production monitoring.

Rolling error rate vs rollback threshold