How our scores behaved before real events
We took events from public lists that we did not make, and checked whether our scores moved in the 60 days before each one. Every result is here, including the misses, next to how often the same scores move that much on any ordinary 60 days.
How this works, in plain words
- The events come from outside. Three public lists, each with a fixed rule written down before we looked at any of our scores: days the VIX first closes above 30 after a calm spell (FRED), the start of each US recession dated by the NBER, and months with a surge in FEMA major disaster declarations (OpenFEMA). The rules and sources are listed below. We do not add or drop events by hand.
- The products are picked by category, not by result. Each event family is matched with every public product in a related catalog category (markets and finance for the VIX, macro and labor for recessions, climate and hazard for disasters).
- One simple test. For each event and product, we compare the product's last real score before the event with its score 60 days earlier. If it moved by 10 points or more, up or down, we call it a hit. Otherwise it is a miss.
- The yardstick. Some scores swing a lot all the time. So for every product we run the same 60-day comparison on every week of its own history. The share of those ordinary windows with a 10-point move is its base rate. If our hit rate is close to the base rate, the moves before events were nothing special.
- Only real history. Synthetic data (stand-in history we show until a real backfill exists) is removed before anything is counted. When a product has no real scores around an event, we say "not enough real history" rather than guess.
Loading the track record…
Headline result
Hit rate before events
Base rate, ordinary 60-day windows
Pairs without enough real history
Synthetic points excluded
This is not evidence that our scores predict events. Our validation scoreboard tests every metric against a naive baseline, and none is yet distinguishable from chance. How we build scores is on the methodology page.
The event lists and their rules
Every event, every product
Open an event to see each product's real score from 90 days before to 30 days after it. Line: the score (0–100). Shaded: the 60 days before the event. Vertical line: the event date. Dots: the two scores compared.
What this does not show
Informational only; not investment, financial or professional advice.