Methodology

A prediction-exchange contract price already is a probability. We do not apply sportsbook de-vigging to it. We surface the displayed price, best bid, best ask, and the bid/ask midpoint — and treat the spread and order-book depth as an uncertainty band, not an edge.

Sportsbook odds carry a bookmaker margin (the vig). We average each outcome's decimal price across surveyed books, invert to implied probability, and normalize the outcome set to sum to 100% — removing the margin. This de-vig engine is the SkipOdds engine, applied only to the sportsbook leg.

Two venues' markets are only merged when they are genuinely the same event — matched on event identity, outcome wording, date ranges, thresholds, and resolution source. Each link is labelled exact / near / related / inverse / conditional, with a confidence score. We never merge on title similarity alone; low-confidence links are flagged for review while still publishing with a label.

The fair probability is a weighted blend of every available source (exchange midpoints, the de-vigged sportsbook consensus, polls, models), weighting each by source type, liquidity, recency, spread, and market quality. The Probability Gap is any single venue's distance from that fair value, in percentage points.

Market price is not the same as reliable probability. The Market Quality Score blends liquidity, bid/ask spread, order-book depth, volume, price stability, time remaining, resolution clarity, source reliability, and cross-platform confirmation. Weights are configurable, not hard-coded in the interface.

Separate from Market Quality, the Confidence Score reflects confidence in the fair probability itself — the number of independent sources, their agreement, recency, reliability, contract-match confidence, and data completeness. Reported as High / Medium / Low / Insufficient Data.

As resolved history accumulates, we measure calibration with Brier score, log loss, and reliability curves — for both the venues and the SkipSeek Fair Probability, by category and platform.

A probability difference is not necessarily profitable after fees, spread, market impact, resolution uncertainty, execution delay, and model uncertainty. We label differences as a Probability Gap, Market Premium, or Market Discount — never as an "edge".

v1 sources are Polymarket (public API) and the SkipOdds de-vigged sportsbook consensus. Additional venues (including Kalshi and CME) are added only under the appropriate data licence; restricted sources are never published until licensed.

SkipSeek publishes probabilities and research for informational purposes. It is not investment or trading advice. Prediction markets may be restricted in your jurisdiction. AI-generated statements are never presented as fact without a cited source.