The term "quantitative research" suggests complexity and opacity — black boxes, impenetrable mathematics, outputs that cannot be explained. In practice, it is something different: a discipline defined by testability, intellectual honesty about its limits, and a direct line between research and portfolio decisions.
What It Actually Is
At its core, quantitative research means applying statistical and mathematical methods to investment questions. Instead of forming a view on a company's management quality or competitive position, you ask: can I measure this? Can I model it? Can I test whether it has predictive content in historical data?
This does not make qualitative factors unimportant. It means working systematically with what is measurable and being honest about the boundaries of your models.
The Research Process
Our research follows a consistent structure:
Observation. Identify a pattern or relationship in market data — a statistical regularity, a risk factor that appears persistent, a structural feature of how a market segment behaves.
Hypothesis. Form a testable explanation for why the pattern might persist. If you cannot articulate a reason, the observation is likely noise.
Testing. Validate against historical data, with realistic assumptions about transaction costs, liquidity, and the constraints that apply to a real portfolio. This is where most ideas fail — not because the pattern was imaginary, but because the costs of acting on it exceed the benefits.
Deployment. If a model passes testing, integrate it into the live process with defined monitoring criteria. The research does not end at deployment — it continues through ongoing performance review.
Where Most Ideas Fail
Step three is the critical filter. Historical data is rich enough that many spurious patterns appear statistically significant. The standard tests for significance were not designed for the scale of modern data — and naive application produces many false positives.
We apply strict out-of-sample testing: the model is evaluated on data it was not trained on, across multiple time periods when possible. We look for consistency rather than peak performance. A model that is robust across different market regimes is more valuable than one that excels in a single favorable period.
The Limits We Acknowledge
No model is complete. Markets evolve, and patterns that held in one regime may not hold in another. Volatility changes. Correlations shift. Structural features of a market can be altered by regulation, technology, or changes in investor composition.
Our models are reviewed regularly. When a model's behavior diverges from expectations in a sustained way, we investigate the cause — we do not retrofit an explanation or ignore the divergence. This intellectual discipline is not a constraint on performance. It is the condition for building something that remains reliable over time.
Research and Operations Are the Same Discipline
The output of quantitative research at KB is not a recommendation or a report. It is a systematic process: one that generates portfolio weights, monitors risk conditions, and identifies when the state of the market warrants a review. Research and portfolio operations are not separate functions — they are the same discipline applied continuously.
This is what it means to run a research-driven fund. The process never stops.
Past performance is not indicative of future results.