◆ EDJE
◆ EXPLAINER · /WHAT-IS-RECONCILIATION

What is reconciliation in prediction markets?

Reconciliation is the process of merging multiple independent signal sources into a single calibrated probability for a market. It's the layer between raw data and a trading decision. Here's what it actually is, why it beats comparison shopping, and how EDJE does it with 47 data sources through a 13-module AI stack.

The short answer

Reconciliation is what happens when you stop asking "what does one source say?" and start asking "when all 47 sources are weighed against each other, what's the most defensible probability?"

Why reconciliation beats comparison

Comparison shopping tools tell you which venue has the best price on a specific outcome. That's useful for arbitrage, but it doesn't tell you whether the price itself is right. A market can be fairly priced on every venue and still be wrong.

Reconciliation doesn't care which venue is cheapest. It asks: given everything we know — expert forecasts, poll aggregators, social sentiment, whale wallet flow, news events, AI ensemble outputs — where should this market actually be priced? Then it compares that number to where it IS priced and surfaces the delta.

How EDJE's reconciliation works

  1. Ingest. 47 independent data sources across 9 tiers — polls, forecasters, markets, social, news, whales, AI ensembles, economic indicators, sector-specific feeds.
  2. Normalize. Each source expresses probability differently (confidence intervals, point estimates, movement vectors). We transform everything to a single probability scale.
  3. Weight. Each source has a learned weight based on its historical accuracy per category. Expert forecasters weight higher in elections; whale flow weights higher in crypto. These weights retrain weekly against closed markets.
  4. Blend. Sources combine through the 13-module calibration pipeline — M1 (median), M2 (weighted mean), M3 (Bayesian update), M4 (calibration-v2), M5–M13 (news-shock detection, residual correction, cross-market consistency, M13 cross-venue price-gap).
  5. Calibrate. Historical Brier scores keep the output honest — if we said 70% and it only happened 60% of the time historically, we discount accordingly.
  6. Surface. Final output: a calibrated probability, the market price, the edge in percentage points, and the reasoning.

Who uses reconciliation?

Try the reconciliation engine

EDJE is live at getedje.com. Free to use for the base picks board. The full Terminal with 10 interactive visualizations is behind a Pro tier.

See live edges →