RESEARCH INFRASTRUCTURE
Research before prediction.
Markets are not a single problem to be solved. They are many overlapping systems — capital flows, fundamentals, positioning, sentiment, liquidity — each moving on its own clock. Our infrastructure is built to hold all of them in view at once.
01
Evidence, then convition.
A forecast is only as good as the research beneath it. Before any view is formed, the system assembles the evidence: what the filings say, what the market has priced, how comparable situations have resolved, and where the current environment sits relative to history.
This ordering is deliberate. Prediction that runs ahead of research produces confident answers to poorly framed questions. We would rather be slower to a view and more certain of its foundation.
02
Point-in-time architecture
Every input the system uses is anchored to what was actually knowable on the date in question. Filings are read as they were filed. Estimates are held as they stood. Prices reflect the tape at that moment, not the tape as later revised.
This discipline is unglamorous and expensive to maintain, and it is the difference between research that describes the past and research that would have survived it. Without it, every backtest is a story told by a narrator who already knows the ending.
03
Breadth and depth
Systematic breadth without fundamental grounding produces patterns that are statistically real and economically meaningless. Fundamental judgment without systematic breadth produces conviction that cannot be tested at scale.
We hold both. Quantitative work identifies where to look and how strongly the evidence points; fundamental analysis asks whether the relationship makes economic sense and what would have to be true for it to persist. A signal that cannot survive both examinations does not enter the process.
04
Dynamic regime awareness
Relationships that hold in one market environment fail in another. Rather than assuming stable behavior across time, the system measures how conditions have actually shifted and adjusts what it expects accordingly.
The goal is not to predict regime changes. It is to avoid the more common and more costly error: applying yesterday's relationships to today's market without noticing the environment has moved.
05
Risk is part of the thesis
Risk management is not a constraint applied after an idea is formed. It is part of forming the idea. Every position carries an explicit account of what would invalidate it — the earnings outcome, the liquidity condition, the correlation that would need to break.
Positions are sized to what can be lost rather than what might be gained, with concentration limits, correlation screening, and event-risk scanning built into the construction process rather than bolted onto it.
06
Thematic depth
Some of the most consequential shifts in markets are structural rather than cyclical — capital reallocating across an entire industrial stack over years, not quarters. These require research that is deeper than a screen and more patient than a trade.
Where we develop a thematic view, it is built from the ground up: the supply chain, the capital commitments, the physical constraints, and the second-order beneficiaries most analysis misses. Depth in a small number of areas beats shallow coverage of many.
07
Differentiated return
The purpose of all of this is narrow: to find return streams that are genuinely differentiated, then to hold them with enough understanding to stay with them when they are uncomfortable.
We hold our own work to institutional forensic standards. Claims are decomposed rather than asserted. Inputs that do not demonstrably improve the process are removed rather than defended. What survives is what we are willing to put capital behind.