In this project supervised by OQAM, the analysts Dag Palmstierna, Denis Zernov and Magnus Drøyvold explore the existence and predictability of overnight reversals in the European high-yield bond market. By using a Random Forest machine learning model, they were able to a predict significant share of overnight reversals in the European high-yield bond market. This supports the hypothesis that such reversals are not only present in equities, but also extend to credit markets.
Translating the model’s signals into a trading strategy, the analysts developed a backtest in which the strategy achieved a cumulative return of 9.64%, exceeding a standard buy-and-hold approach by more than threefold. However, transaction costs and liquidity constraints must be taken into account before drawing real-world conclusions.