Methods
We create proprietary datasets focused specifically on events and then apply a customized machine learning approach for forecasting how markets respond to events. The models are combined with realistic trading assumptions to provide actionable trading signals that allow investors and traders to predict and react to events in the financial markets.
1
FEATURE ENGINEERING
As quants we know that a model can only be as good as its inputs. That's why we focus on data quality and timeliness for all the inputs to our model. We sift through traditional and alternative data sources to identify the best inputs for impactful signals. Our proprietary event dataset is custom built using only relevant features to focus on the nuances of event-based analysis. Every input in our process is scrutinized for accuracy and timeliness to provide our model with superior information.
2
MODEL BUILDING
Our expertise in the financial markets stems from our desire to understand the systematic way in which markets behave. We employ best practices for machine learning technology to build models that capture the non-linear relationships inherent in market dynamics. Our models are trained with adjustments for known risk factors as well as realistic trading costs, including assumptions about implicit and explicit trading costs, so that our signals do not falter when moving from the lab into the wild.
3
SIGNAL CREATION
We combine proprietary event data, market-tested expertise in behavioral responses, and best practices with machine learning methods to create a set of actionable trading signals that deliver verified alpha.