Thursday, February 15, 2024
When I interviewed for a job at GCM 15 years ago, Oguz was one of my interviewers. We found agreement on many aspects of quant modeling, except for my use of neural networks as a part of my strategy: “these models are too over-parametrized, there isn’t enough signal-to-noise ratio in financial data to make them work,” he opined. I still got the job (thanks Oguz), and a few years later, Oguz was proved right when I analyzed live trading data and concluded that my neural networks indeed weren’t adding value and removed them. Fast forward a few years, it was Oguz who pushed his team to use neural networks, now in the context of natural language processing, culminating in a successful news sentiment trading signal. This epitomizes Oguz as a quant researcher: always intellectually honest, and always intellectually curious. When we were trying to evolve trend-following models into the third generation, Oguz built the first adaptive model from his engineering toolbox. When we needed a value signal to complement other components in a new investment product, Oguz took on the project and delivered the results. I am saddened to lose a colleague and top-notch researcher, but Oguz will be fondly remembered through his integrity, work ethic and ingenuity.
