This thesis research focusses on winning at sea. "Winning at sea" requires presence of relevant capability at the right time and for the required time. A winning fleet is a fleet that has the capability to achieve this. The goal of this research is to define a winning fleet and to create a method that can provide insight in how such a fleet can be obtained and maintained. One of the main challenges when operating a naval fleet is to match the fleet capabilities with the capabilities required from the missions that occur over time. The first part of this problem lies in the fact that it is never known what the future will bring. Especially when looking at a period of 30 years which is the general lifetime of a naval vessel. In practice this is partly countered by introducing a major update at the halfway point in the ship's life cycle. The other part of the problem is that the fleet capabilities need to be distributed over the different vessels and vessel types. Because having an infinite fleet size would be far too expensive, certain compromises need to be made. The main of which is that having a finite fleet will result in lower costs but also means that when a vessel is occupied all capabilities of that specific vessel are unavailable for other activities. For this reason it is important that the fleet composition is constructed in such a way that at all times, or at least within an acceptable error, the requirements set by the missions can be met. The method that is created during this research acts like a proof of concept of whether such a method could be useful in the future. This research should be able to help with the iterative process of balancing the operational need and design requirements with the feasibility and affordability that occurs within the early design stages of naval fleet design. The main stakeholders in this process are Commando Zeestrijdkrachten (CZSK), the department of planning (DPLAN), and the Defence Materiel Organisation (DMO). These three parties can respectively be categorized as the user, military planner, and the supplier. In order to reach the research goal a model has been created that simulates naval fleet behaviour. This Fleet Behaviour Model (FBM) describes a fleet as a collection of capabilities which are distributed over different vessel types. In order for this fleet to be operational it needs to comply with certain Life Support Activities (LSA) such as maintenance and the training of the crews. The operational need of this fleet is simulated by generating mission scenarios which require certain capabilities from this fleet for a specific time and at a specific location. Next to requiring mission specific capabilities from the fleet the missions also have the possibility of enemy presence. When this happens naval combat is simulated which introduces attrition into the model. The inclusion of attrition into a fleet behaviour model is what makes this research unique. Using this model the performance of multiple fleets can be tested against one or more scenarios. After the Fleet Behaviour Model was finished and tested a Genetic Algorithm (GA) has been constructed in order to systematically find an optimal solution. In this case an "optimal solution" is the fleet of which the capabilities and the distribution of these capabilities best fit a given mission scenario. It also takes into account the difference in design priorities that can change over time. One or more of three performance parameters: mission successfulness, attrition, and fleet size can be prioritized. Depending on factors like risk and budget these priorities can change which in turn can have a drastic impact on the resulting fleet compositions. In order to test the models' capability to provide insight into what makes an effective fleet, three test cases have been constructed. These test cases test the model on how well it can generate performance optimized fleet compositions as well as testing the performance of one fleet against a range of different scenarios. After these simulations the resulting data has been analysed in an attempt to find clues on what makes one fleet perform better than another. This insight was than translated to the broader spectrum of general early stage naval fleet design. In conclusion, the data gathered from the test cases has proven that the model is capable of providing insight into the early stage design requirements of naval fleet design. As a proof of concept this Fleet Behaviour Model seems promising as a useful tool for naval fleet design in the near future.