Methodology

Market and Model

Market is the indicative quote for the outcome shown on the card. It is a snapshot, not a guaranteed execution price. The price you can actually trade at may differ.

Model is an experimental probability estimate. It is not an observed fact, a validated accuracy score or a promise that an outcome will happen. Markets without a supported model remain visible without a model estimate.

Selecting YES or NO

Cards select the supported outcome whose model estimate exceeds its indicative market price. For a binary YES/NO market, if the model prices YES below the market, we display NO using complementary probabilities. A derived complement price is not a verified ask or executable quote. Both panels, the gap and Market ROI always refer to the selected side. Without a valid, fresh model difference, no favored side is shown.

Gap: a difference, not a return

Gap = (model probability − market price) × 100 percentage points

The selected side has a positive model-minus-market gap. Top picks rank this probability difference, not expected return or guaranteed profitability. Original model evidence remains explicitly labeled with its original outcome.

Market ROI if the selected side wins

Market ROI if the selected side wins = (1 ÷ market price − 1) × 100%

This is the conditional net return if the displayed outcome wins and pays $1 per contract. It uses the market price, not the model probability. For example, buying at $0.495 and receiving $1 gives about +102% net return before costs.

If the outcome loses, the entire stake can be lost. This figure excludes fees and slippage. It is not expected profit and does not account for the chance of losing.

What the models cover

Crypto: supported Bitcoin and Ethereum contracts asking whether the price will be above a threshold at a specified time, up to seven days ahead. The baseline uses spot price and at least 240 closed hourly candles to estimate volatility, with a zero-log-drift Gaussian assumption. It does not model touch, barrier or range contracts.

Sports: supported pregame MLB winner markets, matched to the teams and scheduled game time. The baseline uses current-season win/loss records, smoothing and a neutral matchup calculation. It does not adjust for starting pitchers, injuries, home field, rest or playoffs. Other sports may appear without model estimates.

How the dashboard is organised

• Top picks: up to four fresh, modeled divergences across distinct event groups, with crypto and sports representation when coverage allows.
• Hot markets: up to ten distinct events ranked by the displayed contract's 24-hour activity. Activity does not imply model conviction.
• Weekly whales: public weekly trading-volume leaders across all categories. Volume is not a profitability ranking or evidence of private information.
• WHY: a short explanation using available model inputs and contextual commentary. Trader opinions and AI interpretation remain separate from the numeric model.

Freshness and uncertainty

Quotes older than one hour are excluded from actionable rankings. A snapshot expires after 30 minutes; stale data is reference-only. Individual commentary and leaderboard records also have their own expiry times.

Any sensitivity range shows how an estimate changes when assumptions change. It is not a confidence interval. These models are not calibrated or backed by a published performance track record. Commentary does not automatically change their probabilities.

What Oddictor does not do

Oddictor provides experimental, non-personalized research. It does not execute trades, hold funds or guarantee returns. Coverage is a selected sample, not the entire market. Alerts are currently an owner-only test, not a public subscription service.