The Economist observation: “Recent misses by prediction markets on the outcomes of primary elections in America suggest they may not yet be the ‘global truth machines’ that founders claim.”
Brier Calibration Score
0.424
Mean squared error (0.00 = perfect, >0.25 uncalibrated)
Overconfidence Bias
0.272
Average excess probability assigned to losing outcomes
Calibration Status
Overconfident / Uncalibrated
High odds assigned to candidates who failed to win
Total Races Analyzed
5
Active primary election forecast cohort
Reliability Curve & Outcome Scatter
Preset: 2026 Sample
45° Perfect Calibration
Empirical Market Curve
Loss (Actual 0)
Win (Actual 1)
Forecast Roster & Odds Adjuster
Drag sliders to scrub market implied odds
Candidate / Race
Implied Odds
Adjust Probability
Outcome
Residual Error
Econometric Controls
Primary Election Presets
Select curated historical or simulated primary cohorts based on recent reporting.
Market Stress / Longshot Skew1.00x
Applies non-linear favorite-longshot shrinkage or amplification to implied odds.
Decision Call Threshold0.60
Probability cutoff where market treats a frontrunner as a presumptive winner.
Bayesian Shrinkage Dampening
Re-centers low-liquidity longshot bets toward the historical base-rate mean (0.50).
Audit Documentation
Generate a verifiable statistical calibration report with exact calculations.
Analytical Verdict & Statistical Grounding
The current cohort demonstrates a Brier Score of 0.424 and an Overconfidence Bias of 0.272 across 5 races. Because prediction markets assigned high confidence (e.g. 88% and 65%) to candidates who ultimately lost their primary races, the empirical reliability curve bows significantly below the 45-degree diagonal. This confirms The Economist's finding: prediction markets can suffer from sentiment bubbles and liquidity clustering, diverging from theoretical "global truth machines."