BASF operates one of the largest corporate treasury functions in the world, actively managing currency exposure across global operations. To optimize its EURUSD trading strategy, BASF engaged Multiverse Computing to explore whether quantum and quantum-inspired methods could extract better signals from historical FX data than the classical approaches already in use, and deliver measurably better returns on a real investment horizon.
The Challenge
Currency trading signal generation is a noisy, low-signal problem where classical machine learning models often overfit to historical patterns without generalizing to live market conditions. BASF's existing strategy relied on classical technical indicators, and the challenge was to find a modeling approach that could outperform it consistently across multiple market regimes, using a realistic five-year historical dataset, without introducing impractical computational costs or requiring hardware unavailable in a corporate treasury environment.