Mise en place d'une solution avancée d'optimisation de portefeuilles en vue de son intégration au sein d'une plateforme d'investissement discrétionnaire en ligne

Abstract

Launched in 2015 by Gambit (spin-off from HEC Liège), Birdee is an online discretionary management investment platform that offers the opportunity to invest in thematic and profiled model portfolios. Based on Gambit’s proven algorithms, the managed portfolios are composed exclusively of exchange traded funds. Until now, the validation of investment strategies has been based on backtesting, but is limited to a single optimization per rebalancing date. Recently a new feature supports the manager in testing many different optimization configurations for each single rebalancing date. In this way, the new backtester can select the most efficient one according to a defined criterion called the “backtesting objective”. Our first objective was to validate the correct functioning of the so-called “super-backtester”. Then, we tried to understand how streamlined results could be achieved with this new feature by setting up several initial optimization configurations and then letting the tool select the most appropriate one for each optimization date. We decided to create new configurations by changing the calibration parameters as well as the constraints that the optimal allocation had to respect. A set of determined risk and performance indicators were used to analyze the results. Upon completion, we realized that the selection criteria (backtesting objective) for selecting the configurations and that the used calibration window were doomed to fail. Therefore, based on these observations, the revised objective first proposed two new backtest objectives that would be more adapted to the Birdee optimization and second recommended new ways to determine a proper calibration windows. The last objective was to combine the new features of the backtester with new rebalancing algorithms and to bring out the best suited strategies that could be applied in the future. This work has contributed to providing a decision-making orientation for the different problems and opportunities facing the new start-up. Using a quantitative approach, we managed to analyze the different portfolio performances by applying new investment strategies created thanks to the new portfolio rebalancing algorithms and the new features of the backtester.