A hedge fund trades the cone, not the point

This is an anonymized design-partner scenario; the institution is a hedge fund, the figures are illustrative, and no actor has agreed to anything. We are pre-revenue, building the POC. With that label on it, here is what it looks like when an investment committee retires a point estimate and trades the cone instead.

The point estimate retires

A hedge fund IC that trades on a point estimate — "the name is worth 84" — is trading a bet with no hedge. When the name is at 78, the point has no answer to "what did you expect, and how spread was it?" The cone replaces the point with a distribution: the central scenario, the tail scenarios, the entropy. The IC now trades the shape — sizing to the central path, hedging the tail, and reading the entropy to decide whether the conviction was ever there.

The shape is the decision

A cone is not a more detailed point estimate; it is a different kind of object. The Oracle normalizes the sisters outputs into an Ensemble in exactly one place: probabilities sum to one, scenarios sorted descending, entropy in nats. The IC reads the entropy to know whether the model agrees with itself; reads the modes to know whether there are two regimes; reads the tails to know what the hedge is for. The shape is the decision — the point was a collapse of it, and the collapse threw away the part a hedge needs.

A point estimate is a bet with no hedge. A cone is a bet that tells you the hedge ratio. The IC that retires the point trades the shape.

• The cone carries its own doubt: entropy in nats, modes as regimes, tails as contingencies.
• The IC sizes to the central path, hedges the tail, and reads the entropy to know if the conviction was there.
• Normalization in one place (deino-oracle): sum ≈ 1.0, sorted — the shape is a guarantee, not a sketch.

Why the cone is defensible

A point estimate is defensible only when it happens to be right, which is not a strategy. A cone is defensible because it is a shape that survived the question "what if you are wrong?" — the answer is in the tail, with its probability, and the hedge is sized to it. The manifest records the NodeVersions and agent trace that produced the cone, so the shape is reproducible. The IC is not narrating a trade; it is pointing at a record. We are pre-revenue; the cones are produced; calibration is the loop (Gate 1b) that earns the word "probability," and it has not passed.

The point estimate retires not because the cone is more accurate, but because the cone is the only shape that survives the question "what if you are wrong?" The point has no answer; the cone is the answer.

— design-partner IC note