Estay Dynamics
Track record

Has the board been right?

Every month the board's forecasts are frozen: where it placed each technology, and where the model said it would be 3, 6, 12 and 24 months later under each scenario. When a forecast comes due, it is checked against where the board then placed it, and against the simplest rival, a forecast that nothing moves. The model earns its keep only where it beats standing still.

Scored scenario

How the score is kept

Error is the distance, in points on the 0–100 scale averaged over the two axes, between a forecast and the board's position for that technology in the month the forecast was aimed at. Standing still is the same error for a forecast that every technology stays where it was. Skill is 1 − error ÷ standing-still error: above zero the model did better than assuming nothing changes, below zero it did worse.

Right quadrant is the share of forecasts that put a technology in the class (overheated, proven, underpriced, emerging) it was later in. Right direction checks the forecast move in expectations against the move in the collected attention signals, which do not depend on the analyst; moves under one point are not counted.

The board's later position mixes the analyst scores and the signals. A score not re-reviewed since the forecast was made carries over unchanged, which favours standing still; the "re-reviewed" count shows how many comparisons had a fresh analyst score. Records are kept in data/ledger.json, one per month, never rewritten.