How the research process works
A walkthrough of the weekly workflow: signal generation, manual review, sizing, and exit planning.
The weekly workflow has four steps. Each is deliberately separated so a failure at one stage doesn’t contaminate the others.
1. Signal generation
Quantitative models scan the US equity universe (~3,000 names) every Sunday and surface a ranked list of high-probability setups for the coming week. Signals come from a small, intentionally narrow set of features — price/volume structure, relative strength across sectors, and a few macro filters that flip the model into a defensive posture during regimes that historically perform poorly.
2. Manual review
Every signal that survives ranking gets a human eyeball. The review is not about overruling the model — it’s about checking for the specific kinds of mistakes models reliably make: earnings landmines, recent secondary offerings, illiquidity, or news that obviously breaks the assumption set.
About 30–50% of model signals are rejected at this stage. That’s by design.
3. Position sizing
Surviving ideas get a position size based on a fixed risk budget (a percentage of equity at the planned stop). No idea exceeds the cap. There is no “high conviction = bigger size” knob — the model and the rules size; conviction is what got the idea on the list, not what amplifies it.
4. Exit plan
Every idea ships with an entry zone, a stop, and a target. Exits are mechanical: either the stop or target triggers, or a time-based exit kicks in if neither hits within the planned window.
The point of writing the plan up front is that you can’t argue with it later when you’re emotional.
Why share all this?
Because the value is in the process, not the picks. If you understand how an idea gets generated and graded, you can decide whether the framework suits you — and you have a fighting chance of staying disciplined when individual trades go against you.