Indikora Calibration: Provable accuracy, not marketing
When Indikora says 60%, do we win 60%? We publish our probability calibration and Brier score from real outcomes, and we never hide our losses.
What calibration actually measures
A probability is calibrated when the things called 60% likely happen about 60% of the time. That is a different question from whether a call was right. A forecaster can be right often and still badly calibrated, and can be wrong often and still well calibrated.
Most platforms publish how often they were right. This page publishes whether the numbers they attached to those calls meant anything.
How the figure is produced
Every signal stores the probability assigned to it at the moment it fires, before the outcome is known. When the signal closes at its target or its stop, the outcome is written to the same row. Comparing the two across all closed signals is the whole method, and nothing on this page is entered by hand.
The Brier score
The Brier score is the average squared gap between the probability assigned and what happened, counted as 1 or 0. Lower is better, and 0.25 is what you get by saying 50% to everything.
It is shown because a single accuracy figure hides overconfidence. A forecaster who says 90% and is right 60% of the time looks fine on accuracy and bad on Brier, which is the honest reading.
The losses are in here too
Every closed signal counts, including the ones that hit their stop. A record that lists only the winners is a selection, not a record, and it cannot be checked by anyone outside the company.
What this page does not claim
Calibration says the probabilities are honest. It does not say the signals make money. Those are two separate questions and they are answered separately, from the same ledger.