We follow a strict protocol of research in developing new technology and investment solutions to maintain the integrity of the products we build off of them. The following is our philosophy of our research process.
During Research & Development
“It works in practice, but does it work in theory?”
When we begin to develop a model derived from empirical evidence found in the marketplace, the first question we ask ourselves is, “Why should this model be an accurate representation of reality? What is it founded in?” Unless we can find sound philosophical and theoretical backings for why the model should work in practice, the project is scrapped.
“It works in theory, but does it work in practice?”
Sometimes, the reverse process exists: a model is postulated in theory, but empirical evidence does not support our conclusions. Ultimately, this indicates to us that there is a flaw in our theory, and the project is scrapped.
A model must both be theoretically and empirically defensible.
During Implementation
Ockham’s Razor
In developing our models, we design to include as few “moving parts” as possible. By reducing the complexity of our models to the fewest necessary factors, we reduce the chance of over-optimization to history and increase the chance that the model will continue to work into the future.
Markets Change
It is our belief that market dynamics are consistently changing. On top of trying to reduce the number of “moving parts,” we also try to reduce the number of “fixed constructs” in our models. Fixed constants and values are removed and replaced with adaptive processes, giving us the confidence that models will remain relevant in the future.
Robust in Depth and Breadth
Backtesting is a necessary component of model development. Often, it is where the finger is pointed when over-optimized models break down as markets change. To prevent this risk, we only develop systems that work over a large breadth of securities and a great depth of time. If a model only works for specific securities or only in a specific time-frame, we throw the model out.
By design, our models must be as simple as possible, have the ability to adapt over time, and work over a broad breadth of securities and over a great depth of time.
During Integration
Checks and Balances
Models are wrong. Period. End of story. They are exactly what their name implies: a model of reality. By admitting this fact up front, we confront the hardest part of working with models: how to know when they are wrong what to do when they misbehave. We actively work to try to break out models. We actively work to identify weaknesses and possible flaws. In knowing the short-comings, we can work closely with managers to integrate checks and balances into their investment process so that short-term inaccuracies, or even catastrophic breaks, in the model are not devastating events.
Models are wrong. We actively work to find the weaknesses in ours and work with managers to create checks-and-balances in the investment process.
Summary
We aim to develop models that:
- Are founded in the underlying principles of market dynamics
- Are proven by empirical evidence
- Are as simple as possible
- Are adaptive to changes
- Are robust in their usage
- Can be safely integrated into an investment process using checks-and-balances