Every ParlayIQ market carries two numbers: a market price (the current implied probability) and an AI probability (what our model thinks the contract should resolve at). This page explains where those numbers come from, how they're computed, and how to read the gap between them — the edge — so you can decide whether a YES or NO contract is worth trading.
ParlayIQ is built on three feeds: live odds, market microstructure, and historical parity baselines. Each one serves a different part of the model, and the union is what makes the AI probability useful instead of just decorative.
Live odds come from The Odds API and tie the prediction market price to what sharp sportsbooks are quoting on the underlying game. We refresh prices on demand so the market reflects where the underlying event is being priced right now — not last week.
Market microstructure is the order-flow signal: how much YES volume has been traded, how much NO, how many open positions sit on each side. Imbalance is what tells you whether the current price is being driven by a few whales or by a broad base, which feeds the AI confidence number.
Historical parity baselines anchor the model to long-run sport outcomes — what fraction of NFL markets historically resolve YES, and how home-field advantage skews win rates. This is what stops the model from following a noisy short streak off a cliff.
The model itself is intentionally readable. Every sport starts from a base rate — the long-run share of games that resolve YES under normal conditions. To that base we apply a small home-field adjustment (roughly a 3¢ lift toward the home team), plus a sport-specific variance band that captures how unpredictable the league is week to week.
Variance differs by sport for a reason. NFL games tend to be tighter (lower variance, more predictable); UFC has wider variance because a single finish swings the whole matchup. The model respects that instead of smoothing every league into the same number.
Confidence comes from how close the AI estimate lands to the live market price. When the model and the market agree, confidence is high — the signal is clean. When they diverge by a wide margin, confidence drops because our disagreements are exactly where we have to be most careful.
Yes, the math is simple. That is the point. We trade fair prices, no vig, and an AI signal against live market action — not a black-box neural net you have to take on faith.
Edge is the difference between the AI probability and the current market price, expressed in cents:
edge = ai_probability − market_price
A positive edge means our model thinks the contract is more likely to resolve YES than the market is currently pricing — YES looks underpriced relative to the AI signal. A negative edge flips it: NO is the underpriced side, because the market is more bullish on YES than our model would be.
Edge is a guide, not a guarantee. Pair it with the confidence number — a +8¢ edge at 40% confidence is a lot different beast than a +8¢ edge at 90% confidence. The page you land on after clicking a market shows the same numbers in context, plus any sport-specific brief the model has surfaced.