Analysts

How much will a barrel of oil be worth in 5 minutes? How much will banks be willing to lend for in 2030? This work is about finding patterns and signals across products and markets, using data to validate ideas, brainstorming regularly with colleagues, finding appropriate functional forms, regularly calibrating parameters, and creating and maintaining model code in Python.

How can I monetize the modeled price? When should I be willing to sell and when to buy? What risk am I willing to take on? Potentially everything is important, even how the exchange distributes data, or the order in which our strategy implements various decisions. We test ideas in detailed simulations where we replay the market microsecond by microsecond. We can visualize all market situations using GUIs that are built in-house.

“I identify a few trading situations from the past week that I’d like to see in detail in our GUI. I replay the situation microsecond by microsecond, reviewing our algorithm’s log to see if anything unexpected happened. It looks like our model is giving too much significance to incoming trades in the market. I go through the production model reports and run a parameter recalibration. They look good, even out of sample, and I’d like to merge them into production tomorrow. I send a deployment request to Dev and supervision. I also make a note to evaluate the deployed parameters in a month’s time. I hope I didn’t overfit it. We'll see.”