Case Studies
Many Quant Models Are “Fatally Flawed”
As all firms know, backtesting is not a silver bullet. Marcos Lopez del Prado famously wrote: “Put bluntly, without some knowledge of the causal graph, it is possible (even likely) that an investment strategy will be fatally flawed.”
Overfitting will always be a problem and has historically resulted in underperforming trading models. Training an AI model using backtests is no better. It’s like teaching a person how to take a specific test by giving them all the questions ahead of time, then applauding them for memorizing the answers.
This is the reason quant firms rarely use alternative data: it doesn't know how to look ahead. They're blinded by hindsight bias.
That's why we engineered what we like to call ‘foresight bias.’
Foresight in Focus
Rather than back-test massive datasets only to come up with retrofitted and overfitted solutions, we spent the last five years training our model every single trading day to look ahead, or front-test its predictions.
In other words, Increase Alpha was trained over many years to independently and accurately arrive at new predictions.
This kind of rigorous, methodical front-testing is how we got to over 70% directional accuracy and 92% pure alpha.
See For Yourself
Our approach is simple, durable, proven, and repeatable.
Discover true, uncorrelated alpha with our unique, transparent, and industry-leading model.