Analytical Rigor & Scientific Foundations

Forecasting & Scoring Methodology

Forecast Index is designed on peer-reviewed principles of decision science, probability theory, and collective intelligence. Here we document the exact formulas and rules governing our platform.

The Brier Score (Quadratic Scoring Rule)

Evaluation metric for probabilistic accuracy on resolved binary questions

Formulated by Glenn W. Brier in 1950, the Brier score is a strictly proper scoring rule. Under a strictly proper rule, a forecaster maximizes their expected score if and only if they report their true subjective probability distribution. There is zero incentive to exaggerate or hedge.

// Mathematical Definition

BS = (f - o)²

where f is your submitted forecast probability (0.00 to 1.00), and o is the actual event outcome (1.0 for YES, 0.0 for NO).

Interpreting the Scale (Lower is Better)

0.000

Perfect Foresight

Predicted 100% and occurred, or 0% and did not.

0.250

Random Baseline

Equivalent to predicting 50% on every event without information.

1.000

Complete Error

Predicted 100% certainty on an event that failed to materialize.

Consensus Probability Aggregation

Statistical methods used to derive collective market wisdom

Rather than relying on naive unweighted averages that can be manipulated by malicious outliers or brigade voting, Forecast Index employs a tiered statistical aggregation protocol:

Low Sample Resilience (< 3 forecasters): The system uses the direct arithmetic mean and explicitly marks the question as in a low-data state until the sample matures.
Median Anchoring (3 to 9 forecasters): Uses median aggregation to eliminate single-forecaster manipulation.
Symmetric Trimmed Mean (10+ forecasters): Trims top and bottom 10% extreme distributions before calculating the collective consensus.