This paper describes how the Connoisseur advanced control package has been successfully applied to the broke and retention systems of a high-speed paper machine to provide increased stability and runnability. The paper provides an explanation of the plant testing, identification and modelling techniques used to develop a model-based predictive controller. In addition, details of the controller's performance are presented. Increased competition within the paper industry has led to greater emphasis on continuous improvement of product quality and profit maximisation through optimisation of machine runnability. This demand for continuous improvement has led to the development of on-line instrumentation, allowing key wet-end parameters to be continuously observed. In addition, the introduction of highly effective retention chemicals has provided the means to control retention. The difficulty in the past has been how to "close the loop" to incorporate the new instrumentation and chemicals in a co-ordinated control system. Traditional three term control systems do not have the ability to consider the interactions between wet-end parameters during disturbances such as changes in incoming furnish, broke and recovered fibre flow. By employing a model-based predictive control scheme, the actual multivariable character of the wet-end of the paper machine can be acknowledged and incorporated into the controller design. The model makes it possible for the effects of disturbances in the broke and retention systems to be predicted, allowing the controller to make an effective, measured response. In an actual implementation reported here, this approach is shown to provide significant improvements in stability and runnability, ensuring that the machine returns quickly to a stable running condition after a paper break.