---
title: Champion/challenger framework (video)
description: The champion/challenger framework maintains effective predictive models by providing a dynamic approach to training new models to adapt to the constant changes in real-world conditions.

---

# Champion/challenger framework (video) {: #champion-challenger-framework-video }

Implementing a machine learning model is a complex process; it shouldn't end after deploying a model to production. The constant change in real-world conditions requires a dynamic and adaptable approach to predictive modeling using real-world data. This approach relies on active monitoring supported by the _champion/challenger framework_ to maintain effective predictive models. When you build models with AutoML and deploy the best-performing model, this model becomes the initial _champion_; however, selecting an initial champion is not the end of the process. DataRobot's champion/challenger framework allows you to compare models built with DataRobot and models built outside DataRobot head-to-head and perform model replacement when a challenger is more effective than the current champion.

<hr>

## Add a DataRobot model as a challenger {: #add-a-datarobot-model-as-a-challenger }

This video tutorial demonstrates how to add challenger models to your deployment. The champion/challenger framework allows you to evaluate and compare models safely before making changes to your model in production. With DataRobot, you can add an existing model as a challenger or train a new challenger model with new data.

<div style="position:relative;padding-bottom:56.25%;">
 <iframe style="width:100%;height:100%;position:absolute;left:0px;top:0px;" title="Add a DataRobot Model as a Challenger" frameborder="0" width="100%" height="100%"
 allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" src="https://www.youtube.com/embed/atmu1aKR-iE" allowfullscreen></iframe>
</div>
<br>

<hr>

## Add a non-DataRobot model as a challenger {: #add-a-non-datarobot-model-as-a-challenger }

This video tutorial introduces how to add a model built outside DataRobot (with the Python scikit-learn package) as a custom model challenger in an existing model deployment. You can then evaluate and compare this model with the deployed champion.

<div style="position:relative;padding-bottom:56.25%;">
 <iframe style="width:100%;height:100%;position:absolute;left:0px;top:0px;" title="Add a non-DataRobot Model as a Challenger" frameborder="0" width="100%" height="100%"
 allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" src="https://www.youtube.com/embed/_BQmktobf-s" allowfullscreen></iframe>
</div>
<br>

<hr>

## Model replacement in DataRobot {: #model-replacement-in-datarobot }

DataRobot provides two ways of replacing the model associated with a live deployment. Both options allow you to replace an outdated model without any service interruptions. Your replacement actions are logged automatically in the deployment history. This video tutorial illustrates both model replacement options. Option one is replacing the production model with a new model not associated with the deployment. Option two is replacing the production model with a challenger model associated with the deployment.

<div style="position:relative;padding-bottom:56.25%;">
 <iframe style="width:100%;height:100%;position:absolute;left:0px;top:0px;" title="Model Replacement in DataRobot" frameborder="0" width="100%" height="100%"
 allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" src="https://www.youtube.com/embed/cwElVqGQeho" allowfullscreen></iframe>
</div>
<br>

## Read more {: #read-more }

* [Challengers tab](challengers){ target=_blank }
* [Replace deployed models](deploy-replace){ target=_blank }
