Oddictor's displayed Gap is Model probability minus the indicative Market probability for the same selected outcome. For example, 64% Model minus a 60% displayed Market equals four percentage points. If an independently checked executable ask happened to be $0.62, the separate ask-based difference would be two percentage points before fees; Oddictor has not verified that fill. Neither difference is a realized return or proof that the model is right. A displayed market price may be a midpoint rather than a fillable offer, as Polymarket explains in its price methodology.
Three numbers, three questions
Market: Which selected side and snapshot timestamp does the indicative probability represent? Check actual bid/ask separately: a stale mid-price or a tiny best offer is not a meaningful entry for a larger order. Model: What event definition, inputs and assumptions generated the estimate? A model calibrated on a different event class or old information can look precise and still be wrong. Gap: Is the difference robust to that uncertainty, execution and fees? A positive number means investigate, not buy.
In a separate, explicitly hypothetical purchase of 100 YES shares at a fillable $0.60 ask, gross cost is $60 and settlement pays $100 or $0. This $0.60 ask is not an Oddictor quote or the displayed Market from the opening example. If the true probability were exactly 64%, mathematical expected gross payout would be $64 and expected gross gain $4. If total hypothetical costs were $2, expected net gain would be $2. But 64% is an assumption, not a verified true probability; at 60% the expected gross gain is zero before costs. This arithmetic is not a trade, backtest or statement about Oddictor accuracy. The actual cost structure must be checked per contract against the current official fee guidance.
The card's Market ROI · if the selected side wins is a hypothetical winning-payout ratio computed from the indicative price, not a model forecast or realized net ROI. It excludes fees and does not imply that a position can fill at that price.
Why realized ROI differs
Realized ROI is net realized result divided by capital committed, with a clearly stated convention for partial exits, unsettled positions and fees. One winning position can show high realized ROI even if the original probability estimate was poor. A losing position can occur under a reasonable estimate. Assess many recorded decisions, including losses and unfilled opportunities, before discussing calibration or strategy performance. A larger number of trades does not make the average expected return rise; correlated positions can simply repeat one mistaken view.
A defensible research record logs the initial model and version, publication time, contemporaneous market quote, selected side, expected resolution, proposed size and all applicable costs. After resolution, record the actual result and the exact rule used. If a model changed after news arrived, do not backfill that revised estimate as if it had been available earlier.
A safer interpretation
Treat a gap as a prompt for three independent checks: Does the model address the exact contract? Could a position fill at that price and size? Does the entire portfolio already depend on the same event? Read related-market exposure and fees and liquidity. A dashboard can help organize answers; it cannot manufacture a true probability or guarantee a positive net return.