Blind test

Test yourself against the market. Then see what the Oracle said.

benchmark.lead

How this works

1You see the event and a list of instruments. No outcomes and no Oracle hints at this stage — they never leave the server.
2You pick up to three instruments and call the direction. Your answer is committed server-side before any outcome is read.
3Everything opens at once: the actual abnormal returns, the Oracle's forecast over the same number of positions, and the base rate — the accuracy a random pick would have given.

Only 1 tests so far; the aggregate score appears after ten. Published as it stands, whichever way it points.

Loading events…

What is measured here, and what is not

The outcome is taken from prices, not from our own view of how important a story was: abnormal return against a market model fitted on the 120 sessions before the event, standardised by residual volatility. The significance threshold is a two-sided 5% test.

A single event is an anecdote, not evidence. Aggregate measurements with confidence intervals and base rates are published separately and updated alongside the model.

This material is informational and does not constitute individual investment advice.