This is a public research summary of a framework under live testing. It contains no portfolio recommendations or account-level implementation. Nothing here is financial advice.
Fundamental analysis as translation
The difficult first step in fundamental analysis is often not calculation. It is translating events, context, and business narratives into variables that can be observed, compared, and revised.
“Competition is getting worse,” for example, is not yet an input. It may imply pricing pressure, market-share loss, weaker retention, higher acquisition cost, margin compression, greater capital requirements, or a lower valuation multiple. Those possibilities have different evidence and different consequences.
A useful translation therefore asks:
- What changed in the world or in the business?
- Through which operating driver could that change affect results?
- What evidence would reveal the effect?
- Over what time horizon should it become visible?
- What alternative explanation could produce the same observation?
The quality of later valuation cannot repair a poor translation at the beginning.
Risk-Reward as scenario compression
Risk-Reward compresses a larger body of research into a small set of decision-relevant scenarios, usually Bull, Base, and Bear. Each scenario connects the business model, operating drivers, dependencies, risk paths, plausible outcomes, and a valuation range.
The scenarios are not three arbitrary growth rates. They should represent different causal paths. Their purpose is to expose which assumptions produce which outcomes and where present price sits relative to that range.
Probability enters only after the scenarios are coherent enough to deserve it. A numerical probability attached to a vague narrative creates an appearance of precision without improving the model.
Agentic R-D-G as an updateable structure
Agentic R-D-G, or Agentic R/D Graph, is the working structure used to connect research over time. It organizes events, operating drivers, dependencies, risk paths, scenario prices, and the evidence that updates them.
New evidence should change a specific node, relationship, scenario assumption, or probability rather than merely replacing one story with another.
AI can assist with evidence gathering, comparison, counterarguments, and maintaining the graph. It does not supply accountability: the investor remains responsible for defining the question, judging source quality, choosing uncertainty, and making the decision.
Connection to probability and decisions
Once narratives have been translated and scenarios structured, the framework can connect to the language of probability and risk:
- scenarios form a practical state space;
- valuation outcomes become a distribution rather than a single target;
- confidence can be recorded and calibrated;
- risk exposure can be compared with prospective reward;
- decisions can include sizing, waiting, or taking no action.
This is not an attempt to turn company research into a perfectly specified mathematical model or an automatic order generator. It is a way to make qualitative judgment explicit enough to inspect, update, and connect to risk management.
What live testing is meant to learn
The current test is not simply whether a position makes money. It asks whether the framework identifies important variables, separates drivers from symptoms, responds coherently to evidence, improves uncertainty handling, reduces unstructured reactions, and makes mistakes legible enough to revise the process.
The framework remains provisional. Public updates will focus on changes to method and research structure; detailed trades, forecast logs, and positions remain private.